356 KiB
Comprehension history
2026-10-04T18:28:07Z, PR #215 merged at 5baedb5
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mean | 7.38 | 6.63 | 7.00 | 6.88 | 6.88 |
2026-10-08T12:50:49Z, PR #224 at a58a0a8
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 7 | 6.5 | 7 |
| Mid A | 7.5 | 7 | 7 | 7 | 7 |
| Maintainability senior | 7.5 | 6.5 | 7 | 6.5 | 7 |
| Inherited architect | 7.5 | 7 | 7 | 7 | 7 |
| Mean | 7.38 | 6.75 | 7.00 | 6.75 | 7.00 |
Junior A, hardest first:
skills/larv-review/SKILL.md:77"Architecture": Ratchet, the history move, Fresh, Suspendedskills/larv-review/SKILL.md:51"Round cap", with "Re-entry"bin/loop_claude/cycle/judge_reply.py:90verdict_from,cycle/judge.py:31run_judge,cycle/states.py:69transitionbin/loop_claude/cycle/states.py:9thequoted_answerscounterbin/loop_claude/cycle/archive.py:14archivebin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/loop/reload.py:50exec_if_editedCLAUDE.md:22(Todo items), withbin/rank_todo.py:235check_citations
Mid A, hardest first:
skills/larv-review/SKILL.md:86"Architecture"bin/loop_claude/cycle/states.py:69transition, withcycle/run.py:55stepbin/loop_claude/cycle/worker.py:41worker_argsbin/loop_claude/cycle/judge.py:31run_judgebin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/cycle/archive.py:14archivebin/rank_todo.py:161rank,:408rank_detailsbin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:77"Architecture"skills/larv-review/SKILL.md:56"Round cap" → "Count"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESbin/loop_claude/cycle/states.py:69transition, withcycle/run.py:104question_forbin/loop_claude/cycle/archive.py:14archivebin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/loop/reload.py:50exec_if_editedbin/rank_todo.py:161rank,:408rank_details
Inherited architect, would restructure first:
skills/larv-review/SKILL.md:51"Round cap" and:77"Architecture"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESbin/flat folder: tools, shared modules and repo checksbin/loop_claude/project/checkout.py:32read_utf8bin/loop_claude/loop/main.py:127remaining,cycle/run.py:96ruling
2026-10-08T13:03:40Z, PR #224 at 9ea4f0a
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 7 | 6 | 6.5 |
| Mid A | 7 | 6.5 | 7 | 6.5 | 7 |
| Maintainability senior | 7 | 6.5 | 7 | 6.5 | 6.5 |
| Inherited architect | 7 | 6.5 | 7 | 7 | 7 |
| Mean | 7.00 | 6.50 | 7.00 | 6.50 | 6.75 |
Junior A, hardest first:
skills/larv-review/SKILL.md:77"Architecture", with:51"Round cap"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES,:90verdict_frombin/loop_claude/cycle/archive.py:14archivebin/loop_claude/loop/reload.py:50exec_if_editedbin/rank_todo.py:187read_scoring,:408rank_detailsbin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/cycle/run.py:69workbin/loop-claude:7thesys.pathinsert
Mid A, hardest first:
skills/larv-review/SKILL.md:86"Architecture"skills/larv-review/SKILL.md:51"Round cap" → "Count"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESbin/loop_claude/cycle/archive.py:14archivebin/loop_claude/claude_code/entries.py:28written_by_humanbin/rank_todo.py:408rank_detailsbin/loop_claude/cycle/states.py:69transitionbin/loop_claude/loop/reload.py:50exec_if_edited
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:51"Round cap", with:77"Architecture"bin/loop_claude/cycle/states.py:69transitionbin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES,cycle/judge.py:31run_judgebin/loop_claude/cycle/archive.py:14archivebin/loop_claude/claude_code/entries.py:28written_by_humanbin/rank_todo.py:408rank_detailsbin/loop_claude/loop/reload.py:50exec_if_editedbin/loop_claude/loop/main.py:1the--helpdocstring
Inherited architect, would restructure first:
bin/loop_claude/cycle/judge.py:53judge_promptandjudge_reply.py:12JUDGE_RULES: one home for the judge's promptbin/flat folder: shared modules and testsskills/larv-review/SKILL.md:27§3: bold labels to addressable headingsbin/loop_claude/loop/main.py:5--helpandcycle/retry.py:40: numbers restating constantsbin/loop_claude/project/checkout.py:32read_utf8agents/stability-reviewer.md:3,agents/systems-architect.md:3: house styleworkflow-bench/bench.py:26: not runnable from the repo
2026-10-08T14:28:39Z, PR #225 at 13d8b16
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 7 | 6 | 6.5 |
| Mid A | 7 | 6.5 | 7 | 6 | 6.5 |
| Maintainability senior | 7.5 | 7 | 7.5 | 7 | 7 |
| Inherited architect | 7 | 6.5 | 6.5 | 7 | 7 |
| Mean | 7.13 | 6.50 | 7.00 | 6.50 | 6.75 |
Junior A, hardest first:
skills/larv-review/SKILL.md:79"Architecture": Floor, Ratchet, Fresh, Suspended, and the history paragraph at:95skills/larv-review/SKILL.md:53"Round cap", with:64"Re-entry"bin/loop_claude/cycle/judge.py:31run_judge, withgrounded,verbatimandplainbin/loop_claude/cycle/states.py:69transition, withcycle/run.py:55stepand:104question_forbin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/cycle/archive.py:28archive, withcopy_forandfinished_chunksbin/reading_cost.py:145measure, withbin/shell_reads.py:52segmentsand:67operandsbin/rank_todo.py:408rank_details
Mid A, hardest first:
skills/larv-review/SKILL.md:53"Round cap", with "History", "Ratchet", "Fresh" and "Suspended"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES, withQUOTE_SOURCES,ANSWER_DESTINATIONSandANSWERS_RECORDEDbin/loop_claude/cycle/states.py:69transitionbin/loop_claude/cycle/archive.py:28archive, withfinished_chunksandFINISHED_ATbin/reading_cost.py:145measure, withlocated,relativeandcommit_timebin/shell_reads.py:67operandsbin/loop_claude/claude_code/entries.py:28written_by_humanbin/rank_todo.py:408rank_details
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:79"Architecture"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES, withcycle/worker.py:41worker_argsbin/loop_claude/cycle/judge.py:31run_judgebin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/cycle/archive.py:17FINISHED_AT, withfinished_chunksandcycle/run.py:110closebin/reading_cost.py:145measure, withlocatedbin/shell_reads.py:67operandsbin/rank_todo.py:408rank_details, withread_scoring
Inherited architect, would restructure first:
skills/larv-review/SKILL.md:53"Round cap", and:88"Architecture" from History to Suspendedbin/loop_claude/cycle/archive.py:109finished_chunks, withbin/reading_cost.py:106locatedandbin/loop_claude/project/checkout.py:7checkout_rootbin/flat folder: tools, shared modules and testsbin/loop_claude/project/checkout.py:32read_utf8,claude_code/transcript.py:27render,judge_reply.py:87Verdictbesidestates.py:24rulingbin/loop_claude/loop/main.py:9the recorded-answers rule, besidejudge_reply.py:11ANSWERS_RECORDED
2026-10-08T15:54:46Z, PR #225 at 1fc91a5
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6.5 | 6 | 6.5 |
| Mid A | 7 | 6.5 | 7 | 6.5 | 7 |
| Maintainability senior | 7.5 | 6.5 | 7 | 6.5 | 7 |
| Inherited architect | 7 | 6 | 6.5 | 7 | 6.5 |
| Mean | 7.13 | 6.25 | 6.75 | 6.50 | 6.75 |
Junior A, hardest first:
skills/larv-review/SKILL.md:53"Round cap", through "Re-entry", "Fresh" and "Suspended"bin/loop_claude/cycle/states.py:69transition, withcycle/run.py:55stepbin/reading_cost.py:180measurebin/reading_cost.py:242readings_of, with:171unheldand:236figurebin/loop_claude/cycle/judge.py:31run_judgebin/loop_claude/transcript_archive.py:29archive, withcopy_forandfinished_chunksbin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESbin/rank_todo.py:408rank_details
Mid A, hardest first:
skills/larv-review/SKILL.md:53"Round cap", with "Architecture" from Floor to Suspendedbin/loop_claude/cycle/judge.py:31run_judgebin/loop_claude/cycle/states.py:69transitionbin/reading_cost.py:180measure, and:242readings_ofbin/loop_claude/transcript_archive.py:29archive,:95copy_forand:110finished_chunksbin/loop_claude/claude_code/entries.py:28written_by_humanbin/loop_claude/loop/reload.py:50exec_if_editedbin/rank_todo.py:408rank_details
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:79"Architecture"skills/larv-review/SKILL.md:95the move of an old entry's scores into the historybin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESbin/loop_claude/cycle/judge.py:31run_judgebin/reading_cost.py:180measure, and:242readings_ofbin/loop_claude/transcript_archive.py:18FINISHED_AT, withfinished_chunksandcycle/run.py:110closebin/reading_cost.py:141located, besidebin/loop_claude/project/checkout.py:7checkout_rootbin/shell_reads.py:52segments, and:67operands
Inherited architect, would restructure first:
bin/by kind: tools, shared modules and testsbin/loop_claude/transcript_archive.py:110finished_chunks, read bybin/reading_cost.pythrough fourloop_claudemodulesdocs/comprehension-history.md, written bybin/reading_cost.py:242and byskills/larv-review/SKILL.md:92's proseskills/larv-review/SKILL.md:53"Round cap"bin/loop_claude/cycle/judge_reply.py:87Verdictbesidestates.pyruling, andproject/checkout.py:32read_utf8
2026-10-08T17:29:52Z, PR #225 at 8c9bf09
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 7 | 6.5 | 6.5 |
| Mid A | 7 | 7 | 7 | 6.5 | 7 |
| Maintainability senior | 7.5 | 6.5 | 7 | 7 | 7 |
| Inherited architect | 7.5 | 7 | 7 | 7.5 | 7 |
| Mean | 7.25 | 6.75 | 7.00 | 6.88 | 6.88 |
Junior A, hardest first:
bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES, withReplyandverdict_frombin/loop_claude/cycle/judge.py:31run_judge, withgrounded,verbatimandplainbin/loop_claude/cycle/states.py:69transitionbin/loop_claude/transcript_archive.py:30archive,copy_forandfinished_chunksbin/reading_cost/measure.py:27measure, withbin/reading_cost/main.py:113locatedskills/larv-review/SKILL.md:95the old entry's scores, and:108"Fresh"bin/reading_cost/shell_reads.py:21shell_reads, withoperands,segmentsandshell_linesbin/loop_claude/claude_code/entries.py:28written_by_human
Mid A, hardest first:
skills/larv-review/SKILL.md:53"Round cap", through "Architecture", "Fresh" and "Suspended"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES, and:90verdict_frombin/loop_claude/cycle/states.py:69transition, withcycle/run.py:35run_cycleand:104question_forbin/loop_claude/transcript_archive.py:30archive,copy_forandfinished_chunksbin/loop_claude/claude_code/entries.py:28written_by_humanbin/reading_cost/figures.py:34readings_ofbin/reading_cost/shell_reads.py:52segments, and:67operandsbin/reading_cost/main.py:113located, besidebin/loop_claude/project/checkout.py:7checkout_root
Maintainability senior, hardest first:
bin/loop_claude/transcript_archive.py:30archive, and:111finished_chunksbin/reading_cost/figures.py:34readings_ofbin/reading_cost/shell_reads.py:52segments, and:67operandsbin/loop_claude/cycle/judge.py:31run_judgebin/reading_cost/main.py:113located, andmeasure.py:21Checkoutskills/larv-review/SKILL.md:53"Round cap", and:108"Fresh" and "Suspended"bin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESbin/rank_todo.py:408rank_details
Inherited architect, would restructure first:
bin/loop_claude/claude_code/and the read side ofbin/loop_claude/transcript_archive.py, into a package both tools importskills/larv-review/SKILL.md:58"Count", with a checkerbin/loop_claude/cycle/judge_reply.py:12JUDGE_RULES, besidejudge.py:53judge_promptbin/loop_claude/project/checkout.py:7checkout_rootandbin/reading_cost/main.py:113located, andcheckout.py:32read_utf8bin/failure.py,bin/file_writes.pyandbin/markdown_sections.py, into one package
2026-10-08T18:37:21Z, PR #228 at 2682668, against 8c9bf09
Ruling: worse
| Seat | Vote |
|---|---|
| Mid A | worse |
| Inherited architect | worse |
Mid A, decided by:
skills/larv-review/SKILL.md:91"Paired ruling": the base, its fallbacks, a two-stage vote and a tie rule in one paragraphskills/comprehension-panel/scoring-run.md:1and:67"Paired ruling": one file name over two kinds of promptskills/comprehension-panel/SKILL.md:102"The paired ruling": the procedure spread over three filesskills/larv-review/SKILL.md:89"Architecture": the comprehension check, paired ruling, scoring run, ratchet and floor to hold at onceskills/larv-review/SKILL.md:105"History": two record shapes in one bulletskills/larv-review/SKILL.md:66"Re-entry", better: "Rerun" names its mechanism
Inherited architect, decided by:
skills/larv-review/SKILL.md:89"Architecture": two mechanisms, each with its own trigger and seats, and a three-way rule for the baseskills/larv-review/SKILL.md:105"History": two entry shapes, and which commit of a paired entry is the next baseCHANGELOG.md:5against:11,:16,:19: two accounts of what blocks a mergeCLAUDE.md:37and:40"Quality": "comprehension check" undefined, and one sentence packing the floor and the rulingskills/comprehension-panel/scoring-run.md:1"Paired ruling": the name no longer says what is insideAGENTS.md:15: the new decision title packs both mechanisms and the vote ruleskills/larv-review/SKILL.md:35and:64"Re-entry", better: a rerun traces from the table alone
2026-10-08T18:46:12Z, PR #228 at fbe753e, against c7333c0
Ruling: same
| Seat | Vote |
|---|---|
| Mid A | worse |
| Inherited architect | better |
| Maintainability senior | same |
| Junior A | worse |
Mid A, decided by:
skills/larv-review/SKILL.md:89"Architecture": two kinds of run under one umbrella, each with its own trigger, and a vote rule to work through by exampleskills/larv-review/SKILL.md:93"Base": the comparison point found by how earlier chunks appended to the historyskills/larv-review/SKILL.md:35the Architecture row: a forward reference to the scoring run's triggerCLAUDE.md:37and:40"Quality": one concept under several names, and "the last measured commit" undefinedskills/comprehension-panel/SKILL.md:102"The paired ruling": who votes is left to the callerskills/comprehension-panel/seat-prompts.md:1, better: the name tells the truth, and{forbidden}serves both briefsskills/larv-review/SKILL.md:64"Re-entry", better: the rerun rule is explicit
Inherited architect, decided by:
skills/larv-review/SKILL.md:64"Re-entry", better: why a pass ran a fourth time is answered in one paragraphskills/larv-review/SKILL.md:33the pass table, better: each pass's rule for fixes sits in its own Trigger cellskills/larv-review/SKILL.md:37and:123"Security", better: when each security pass runsskills/larv-review/SKILL.md:93"Base": base selection depends at a distance on the wording of a line in the history formatskills/larv-review/SKILL.md:35the Architecture row: a pointer 60 lines downskills/larv-review/SKILL.md:95"Ruling": the vote margin worked out by handCLAUDE.md:37andAGENTS.md:16: one concept, three names
Maintainability senior, decided by:
skills/larv-review/SKILL.md:89"Paired ruling": the base defined by an action, and the vote arithmeticskills/larv-review/SKILL.md:35and:97: a forward reference, though one definitionskills/larv-review/SKILL.md:64"Re-entry", better: deterministic rerunsskills/larv-review/SKILL.md:59"Later rounds", better: every rerun reviews only its diffskills/comprehension-panel/seat-prompts.md:3, better: the briefs cannot drift apartAGENTS.md:35"Scoring run": the heading no longer matches its useCLAUDE.md:40: "last measured commit" defined only in larv-review
Junior A, decided by:
skills/larv-review/SKILL.md:91"Paired ruling": the base, the escalation and the vote rule in three rereadsskills/larv-review/SKILL.md:35the Architecture row: a forward reference to a two-condition ruleskills/larv-review/SKILL.md:35,:51,:89,CLAUDE.md:37: four names for the comprehension machineryskills/larv-review/SKILL.md:36the Full scan and Security rows: read together nowskills/comprehension-panel/seat-prompts.md:8: placeholders nesting placeholders
2026-10-08T18:57:27Z, PR #228 at 50249b1, against 12d6d75
Ruling: worse
| Seat | Vote |
|---|---|
| Mid A | worse |
| Inherited architect | worse |
Mid A, decided by:
skills/larv-review/SKILL.md:93"Base": agit logcommand, a two-way condition, a forward reference to the history's format and two fallbacksskills/larv-review/SKILL.md:89"Architecture": two kinds of run, four sub-bullets and a vote rule; the section's fan-out roughly doubledskills/larv-review/SKILL.md:35,:98and:120: one trigger written out three timesCLAUDE.md:40"Quality": "the last commit the panel scored or ruled better" reads two waysAGENTS.md:35"Scoring run": the heading no longer names what it feedsCHANGELOG.md:5: ten sentences in one entryskills/larv-review/SKILL.md:66"Re-entry",:37and:124"Security",skills/comprehension-panel/seat-prompts.md:7, better
Inherited architect, decided by:
skills/larv-review/SKILL.md:93"Base": finding the base at 3am takes git archaeology across merge commitsCHANGELOG.md:5againstskills/larv-review/SKILL.md:93and:107: only the changelog says a ruling of same leaves the base, so a chunk may be told to fix places it never touchedskills/larv-review/SKILL.md:95"Seats" and:96"Ruling": the vote table built in the headCLAUDE.md:40"Quality": a folded noun phrase in the charterskills/larv-review/SKILL.md:35,:98and:120: the trigger three times, and seven concepts where there were threeAGENTS.md:35"Scoring run": the heading not renamed with the fileskills/larv-review/SKILL.md:64"Re-entry" and:40the Prose row,:37and:124"Security", better
2026-10-08T19:48:05Z, PR #229 at 8a2dbd7, against 12d6d75
Ruling: worse
| Seat | Vote |
|---|---|
| Mid A | worse |
| Inherited architect | worse |
Mid A, decided by:
skills/larv-review/SKILL.md:93"Paired ruling" → "Base": the base found by a prosegit logsearch with two fallbacksskills/larv-review/SKILL.md:95"Seats" and "Ruling": vote arithmetic written as prose, with no exampleskills/larv-review/SKILL.md:89"Architecture": two kinds of run, each with its own trigger, seats and blockCLAUDE.md:40"Quality": a clause folded into a noun phrasebin/metric_bench/places.py:7RUN,SEAT_PLACES,PLACEand:30scoring_runs: a second reader of a format defined only in prose, linked from neither side
Inherited architect, decided by:
skills/larv-review/SKILL.md:93"Paired ruling" → "Base": "worse than what?" means running the base search in your headskills/larv-review/SKILL.md:95"Seats" and "Ruling": a two-stage panel and vote arithmetic with nothing to run or test itskills/larv-review/SKILL.md:35,:98"Scoring run" and:120"Suspended": the scoring run's trigger written out in four placesskills/larv-review/SKILL.md:36the Full scan and Security rows: each row's trigger depends on the otherbin/metric_bench/places.py:7RUN,SEAT_PLACES,PLACE: a regex reader of the history linked from neither sidebin/metric_bench/places.py:59resolved: a result that depends on the place resolved before itbin/metric_bench/python_units.py:69Cognitive: nesting bookkeeping spread over its methods
2026-10-08T21:14:15Z, PR #229 at 5cc37dd, against 12d6d75
Ruling: worse
| Seat | Vote |
|---|---|
| Mid A | worse |
| Inherited architect | worse |
Mid A, decided by:
skills/larv-review/SKILL.md:93"Paired ruling" → "Base": a rule over git history, agit logcommand and two fallbacks to hold at onceCLAUDE.md:40"Quality": a garden-path sentence read three timesskills/larv-review/SKILL.md:89"Architecture": two kinds of run, each with its own trigger, meeting "Fresh", "Suspended" and "Round cap"bin/metric_bench/places.py:60resolved: order-dependent, readable only with the history format in mindbin/metric_bench/main.py:43main, better: the shape ofbin/reading_cost/main.py, one for oneskills/larv-review/SKILL.md:111"History", better: the format and its reader name each other
Inherited architect, decided by:
skills/larv-review/SKILL.md:93"Paired ruling" → "Base": the ratchet's outcome rests on state spread across many past commitsskills/larv-review/SKILL.md:89"Architecture": one mechanism became two, with different triggers and blocksCLAUDE.md:40"Quality": a clause folded into a noun phrasebin/metric_bench/places.py:7RUN,SEAT_PLACES,PLACEand:60resolved: a regex reader of a prose format, and a historical quirk to know firstAGENTS.md:35"Scoring run": the section's name no longer covers its use by paired rulingsskills/larv-review/SKILL.md:64"Re-entry", better: which passes a fix reruns is explicit
2026-10-08T23:12:17Z, PR #231 at 53ed515, against 12d6d75
Ruling: better
| Seat | Vote |
|---|---|
| Mid A | better |
| Inherited architect | better |
Mid A, decided by:
bin/larv_review/rules.py:64round_counts, with:10CAPandbin/larv_review/main.py:118round_count, better: a fourth round leads cold to one named function, where the base had onlyskills/larv-review/SKILL.md:58"Count" to carry out by handbin/larv_review/main.py:1the module docstring, better: the base, round count, trigger, vote rule and seats file are written beside the code that carries them outbin/larv_review/rules.py:41rulingand:27votes, better: the vote rule is a short pure function, changeable without the skill text in mindbin/larv_review/git.py:4the import ofHistoryChangefrom rules, with:50history_changes: the git layer takes its result type from the rules module, so the direction among the three modules is not obviousbin/larv_review/history.py:85resolvedand:56scoring_runs: they serve only metric-bench, so old-history path logic sits in the review tool's modulebin/metric_bench/main.py:143gitagainstbin/larv_review/git.py:9git: two private git runners with different error handlingbin/larv_review/rules.py:106scoring_run_due: the densest new unit, two plans, shipped IDs and a flag held at onceskills/larv-review/SKILL.md:93"Paired ruling" and:100"Scoring run": more vocabulary, though each mechanical step now names a subcommand
Inherited architect, decided by:
bin/larv_review/rules.py:64round_counts, withbin/larv_review/main.py:118round_count, better: the 3am question leads toround-countand one pure function, where the base'sskills/larv-review/SKILL.md:58"Count" was prose to carry out againstgit logbin/larv_review/rules.py:106scoring_run_due, better: the trigger is a checkable function that prints its reason, though dense in itsand/orbinding and set arithmeticbin/larv_review/main.py:1the module docstring, better: the history format and ratchet have one tested home;skills/larv-review/SKILL.md:42now points at--helpfor the rulesbin/larv_review/history.py:85resolved,:56scoring_runsand:61seat_places: a parser only metric-bench uses, under the review tool's name; where it rots firstbin/larv_review/git.py:4the import ofHistoryChangefrom rules, andbin/metric_bench/main.py:143git: a reversed dependency and two git runnersbin/larv_review/seats.py:7SCORING_SEATSandSEAT_LINE, withbin/larv_review/rules.py:11FIRST_SEATS: seat names spelled in four placesbin/larv_review/main.py:164recordandbin/larv_review/history.py:52entries: positional regex groups that needHEADINGopen beside them
2026-10-08T23:40:09Z, PR #231 at efed78b, against 12d6d75
Ruling: worse
| Seat | Vote |
|---|---|
| Mid A | worse |
| Inherited architect | worse |
Mid A, decided by:
bin/larv_review/rules.py:105scoring_run_due: set arithmetic over two plans, anearliestrelease and alandedflag built ingit.py, withorandandunparenthesised, where the base's trigger was one clause atskills/larv-review/SKILL.md:35bin/larv_review/rules.py:66measured_commitand:82base: the base comes from diffing the history around each merged change, held withhistory.pyentriesandgit.pyhistory_changesat oncebin/larv_review/rules.py:16votesand:34ruling, withbin/larv_review/seats.py:16FIRST_SEATS: fixed seat order, a split, a lead of two and a placeless vote stacked, whereskills/larv-review/SKILL.md:93only points at--helpbin/larv_review/main.py:220read_seats:named.group(3, 4)depends onHEADING's group positions inbin/larv_review/history.py:13bin/larv_review/history.py:95resolvedand:66scoring_runs: used only bybin/metric_bench/main.py:40, so metric-bench reaches into another tool's packagebin/larv_review/rules.py:57round_counts, better: the round count is a small named function where the base had prose atskills/larv-review/SKILL.md:58bin/larv_review/history.py:129ruling_entry, better: the history's format has one owner in code
Inherited architect, decided by:
bin/larv_review/rules.py:105scoring_run_due: set arithmetic whose inputs are built in three other files, where the base's trigger was one table cell atskills/larv-review/SKILL.md:35bin/larv_review/rules.py:66measured_commitand:82base:against is Nonedoubles as the type tag ofbin/larv_review/history.py:30Entry; the hardest unit at 3ambin/larv_review/main.py:220read_seats, withbin/larv_review/history.py:111heading: a rawre.Matchwhose groups another file reads by positionbin/larv_review/seats.py:10MEAN, withbin/larv_review/history.py:111heading: a fake seat standing for a legacy import, threaded through every scoring pathbin/larv_review/history.py:95resolvedand:66scoring_runs: metric-bench's legacy path rules inside larv-review's package, and twogithelpers atbin/metric_bench/main.py:143andbin/larv_review/git.py:10bin/larv_review/main.py:1the module docstring, withbin/larv_review/rules.py:83base: the base rule stated in several places in different wordsbin/larv_review/rules.py:57round_counts, better: the fourth-round symptom leads to a tested function
2026-10-08T23:48:44Z, PR #231 at c1133ed, against 12d6d75
Ruling: same
| Seat | Vote |
|---|---|
| Mid A | better |
| Inherited architect | worse |
| Maintainability senior | better |
| Junior A | worse |
Mid A, decided by:
bin/larv_review/rules.py:57round_counts, withbin/larv_review/main.py:141round_count, better: the fourth-round symptom leads to a short function, where the base had onlyskills/larv-review/SKILL.md:58"Count"bin/larv_review/main.py:1the module docstring, better: each rule stated once beside the code that enforces it, where the base hadskills/larv-review/SKILL.md:92"History"bin/larv_review/history.py:12HEADING,:58entriesand:128entry_text, better: the history is read and written in one modulebin/larv_review/main.py:195record_scoring, withbin/larv_review/seats.py:37reportandbin/larv_review/history.py:93heading: the moved Mean path threads through three filesbin/larv_review/rules.py:105scoring_run_due,:119emptiedand:66measured_commit: dense rules that lean on long docstringsbin/metric_bench/main.py:38andbin/larv_review/git.py:5: imports across packages
Inherited architect, decided by:
bin/metric_bench/main.py:38andbin/metric_bench/bench.py:7, withbin/larv_review/history.py:22Placeand:69seat_places: metric-bench reads its input through larv-review's history modulebin/larv_review/rules.py:66measured_commit,:82baseandbin/larv_review/main.py:124ruling_base: three fallbacks to hold at once, where the base's ratchet was one sentence atskills/larv-review/SKILL.md:100bin/larv_review/rules.py:105scoring_run_dueand:119emptied, withbin/larv_review/git.py:88landed_since_release: the trigger spread over three files and the helpbin/larv_review/rules.py:7the imports, withbin/rank_todo.py:386parse_row: rank_todo is now a library for larv-review toobin/larv_review/seats.py:30Reportand:37report, andbin/metric_bench/python_units.py:20measured: names that cost a re-readbin/larv_review/rules.py:57round_counts, better: the 3am question answerable in code
Maintainability senior, decided by:
bin/larv_review/rules.py:57round_counts, withbin/larv_review/main.py:141round_count, better: the 3am symptom lands on a tested unit, where the base hadskills/larv-review/SKILL.md:58"Count"bin/larv_review/main.py:1the module docstring, better: the base search, round count, trigger and seats file stated once, where the base hadskills/larv-review/SKILL.md:92"History" and:100"Ratchet"bin/larv_review/rules.py:66measured_commit, withbin/larv_review/git.py:66history_changes: the costliest new unit to holdbin/larv_review/rules.py:105scoring_run_dueand:119emptied: exact and checkable, but read twicebin/larv_review/rules.py:1the module: six concerns, andvotesdepends onPLACEfrombin/larv_review/history.py:14bin/larv_review/main.py:195record_scoring: two inputs threaded through sentinels
Junior A, decided by:
bin/larv_review/main.py:195record_scoring, withbin/larv_review/seats.py:44reportandbin/larv_review/history.py:93heading: two modes woven through one function and three filesbin/larv_review/rules.py:105scoring_run_due, with:119emptied: needs the release rules fromCLAUDE.mdandbin/larv_review/git.py:88landed_since_releasebin/larv_review/rules.py:66measured_commit:orandandmixed without parentheses, and a new concept to learn firstbin/larv_review/git.py:5andbin/metric_bench/main.py:38: the git layer depends on the history type, and metric-bench on larv-review's packagebin/larv_review/main.py:1the module docstring: one home, but a wall of rules leaning on terms defined elsewherebin/larv_review/rules.py:16votes: a seat protocol reconstructed only fromskills/larv-review/SKILL.md:97
2026-10-09T08:54:36Z, PR #232 at 2c6e063, against 12d6d75
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | same | better | better |
| Inherited architect | better | worse | same | better | better |
Mid A, decided by:
bin/larv_review/main.py:146round_count: Navigation, better: the fourth-round symptom leads to a named unit, where the base had onlyskills/larv-review/SKILL.md:58"Count"bin/larv_review/rules.py:42ruling,:52fallenand:128scoring_run_due: Self-sufficiency, better: vote arithmetic and the scoring trigger are pure functions with docstrings, where the base had prose aloneREADME.md:68"Tools": Navigation, better: the README names larv-review and metric-benchbin/larv_review/main.py:1the module docstring: Self-sufficiency, better, with a cost: Split, Fell, Ratchet and the moved Mean line arrive in long sentencesbin/larv_review/history.py:96heading: Locality, against: two flags with DATED and HEADING, traced throughbin/larv_review/seats.py:38reportandbin/larv_review/main.py:207record_scoringbin/larv_review/history.py:18SEAT_PLACES,:114ruling_entryand:125scoring_entry: Locality, against: writers and parser must agree on label stringsbin/larv_review/rules.py:1the module: Shape, against: one vague name over five concernsbin/metric_bench/bench.py:87resolvedandbin/metric_bench/main.py:38: Shape and Locality, against: metric-bench depends on larv-review's history parser
Inherited architect, decided by:
bin/larv_review/main.py:146round_count: Navigation: the 3am question lands on a function, where the base hadskills/larv-review/SKILL.md:58"Count"bin/larv_review/rules.py:89measured_commitand:105base: Self-sufficiency: the ratchet's base is executable, where the base hadskills/larv-review/SKILL.md:100bin/larv_review/history.py:60entries,:114ruling_entryand:125scoring_entry: Self-sufficiency and Navigation: the history's format is a regex plus writers, where the base hadskills/larv-review/SKILL.md:92bin/larv_review/main.py:207record_scoring, withbin/larv_review/seats.py:45reportandbin/larv_review/history.py:96heading: Locality: the moved Mean case threads through three modulesbin/larv_review/rules.py:57pending_fall, called frombin/larv_review/main.py:179and:224: Locality: commands depend on state an earlier command appended to the history;history[1]hides it at the call sitebin/metric_bench/main.py:38andbin/metric_bench/bench.py:7: Locality: metric-bench parses the history with larv-review'sentries, and keeps its owngit()atbin/metric_bench/main.py:143bin/larv_review/rules.py:1the module: Shape: a grab-bag namebin/larv_review/git.py:56file_atandbin/larv_review/main.py:247history_text: Locality: two histories, the default branch's and the checkout'sskills/larv-review/SKILL.md:97"Seats": Self-sufficiency, against the skill: it needs the tool's help open beside itbin/rank_todo.py:386parse_row: Shape, a small gain
2026-10-09T09:35:29Z, PR #232 at 6a8bbcf, against 12d6d75
Ruling: same
Fell: Locality, Shape
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | worse | same | better | better |
| Inherited architect | better | worse | worse | same | worse |
| Maintainability senior | better | worse | worse | better | same |
| Junior A | better | worse | worse | better | worse |
Mid A, decided by:
- Navigation, Self-sufficiency, Overall:
bin/larv_review/rules.py:118round_counts, withbin/larv_review/main.py:155round_count: the fourth-round symptom leads from the README's Tools list to a named, tested function, where the base hadskills/larv-review/SKILL.md:58"Count" - Self-sufficiency, Overall:
bin/larv_review/history.py:117ruling_entryand:130scoring_entry: the history format is produced by code and checked by a regex, where the base hadskills/larv-review/SKILL.md:92"History" - Locality:
bin/larv_review/main.py:234record_scoring: five conditionals ongiven.moved_meanweave a one-off migration into the main scoring path - Locality:
bin/metric_bench/main.py:38andbin/larv_review/history.py:31Entry: places are parsed in larv_review but used only by metric_bench;bin/metric_bench/bench.py:87resolvedcopes with old paths - Locality:
bin/larv_review/main.py:1the module docstring: rules compressed into one dense block; the fall, Ratchet and Still standing chain reaches acrossbin/larv_review/rules.py:95fall_outcome,:76standing_ratchetandbin/larv_review/main.py:164needs_scoring_run - Navigation:
README.md:66"Tools": lists larv-review and metric-bench;skills/larv-review/SKILL.md:46points at--help - Shape:
bin/larv_review/rules.py:155plan_itemsand:169scoring_run_due: a catch-all module, offset by the clean package split and the truthfulskills/comprehension-panel/seat-prompts.md:1name
Inherited architect, decided by:
- Locality:
bin/larv_review/main.py:234record_scoring: whether a run is allowed and whether the floor blocks depend on the plan diff, a pending fall read from the history, andmoved_mean, checked five times in 25 lines - Locality:
bin/larv_review/rules.py:71pending_falland:76standing_ratchet: commands talk only through the history file, so the outcome depends on the order they ran in - Locality, Shape:
bin/larv_review/rules.py:127measured_commit: re-parses the history around every merged change and applies a two-clause rule; where it rots first - Shape:
bin/larv_review/rules.py:1the module: vote counting, round counting, base search, plan parsing and release emptying under a name that says as little as utils;HistoryChangedefined in history but built inbin/larv_review/git.py:66history_changes - Shape:
bin/metric_bench/main.py:38: a second tool no review runs, depending on larv_review's history parser;bin/gains two packages - Shape:
bin/metric_bench/bench.py:87resolved: a shim for old path formats - Navigation:
bin/larv_review/rules.py:118round_countsandbin/larv_review/main.py:155round_count: the 3am symptom lands on a named command, where the base hadskills/larv-review/SKILL.md:58"Count" - Navigation:
README.md:66"Tools": lists every tool - Navigation:
skills/comprehension-panel/seat-prompts.md:1: a truthful name - Self-sufficiency:
bin/larv_review/main.py:1the module docstring andskills/larv-review/SKILL.md:93"Paired ruling": the code explains itself, but the skill's ratchet text needs the docstring open beside it; ruled same
Maintainability senior, decided by:
- Locality:
bin/larv_review/main.py:234record_scoring: testsgiven.moved_meanfive times, each branch tied tochunk_history,due_reasonsandpending_fall - Locality:
bin/larv_review/rules.py:71pending_fall: its docstring holds only becausebin/larv_review/history.py:62entriesgives scoring runs an emptyfell; the Fell, Ratchet and Still standing chain spans four files - Locality, Shape:
bin/metric_bench/main.py:38main: metric-bench imports larv_review's history internals;bin/metric_bench/main.py:143gitrepeatsbin/larv_review/git.py:10git - Shape:
bin/larv_review/rules.py:35counted: a function and a parameter share the name; the module name says nothing of its contents - Shape:
bin/larv_review/main.py:177due_reasons: main holds commands plus I/O helpers, and the vocabulary grew to seven concepts - Navigation:
README.md:66"Tools": every tool listed with its--help - Self-sufficiency, Navigation:
bin/larv_review/rules.py:118round_counts: the round count, vote count and trigger state their rules in code, where the base hadskills/larv-review/SKILL.md:58"Count" - Self-sufficiency:
bin/larv_review/main.py:1the module docstring: an 80-line block of every rule at once, partly cancelling the gain
Junior A, decided by:
- Locality, Overall:
bin/larv_review/main.py:234record_scoring: the moved-Mean case is checked five times and decides four things - Locality, Overall:
bin/larv_review/rules.py:76standing_ratchetand:71pending_fall: a merge block depends on state earlier commands wrote into the history - Locality, Shape:
bin/larv_review/rules.py:127measured_commit: walks every merged change's history; restated atbin/larv_review/main.py:11, the hardest sentence in the repo - Shape:
bin/larv_review/rules.py:1the module: a grab-bag name beside truthfully named siblings - Shape:
bin/larv_review/seats.py:31Report: the fieldreadlooks like a verb;said,counted,givenandfoundatbin/larv_review/main.py:217record_rulingneed tracing - Shape:
bin/metric_bench/main.py:38: the history format is a contract between two tool packages - Navigation, Self-sufficiency:
bin/larv_review/rules.py:118round_counts: the 3am question lands on one named function, where the base hadskills/larv-review/SKILL.md:58"Count" - Navigation:
README.md:66"Tools": the README lists larv-review and metric-bench - Self-sufficiency:
bin/larv_review/history.py:62entriesand:117ruling_entry: the history format is enforced in code; the skill needs--helpopen beside it
2026-10-09T09:46:21Z, PR #232 at 324d650, against 12d6d75
Ruling: same
Fell: Locality
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | worse | worse | better | same |
| Inherited architect | better | worse | worse | better | better |
| Maintainability senior | better | worse | same | better | better |
| Junior A | better | worse | same | same | same |
Mid A, decided by:
- Navigation, Self-sufficiency:
bin/larv_review/rules.py:125round_counts: the fourth-round symptom leads to a named function andCAP, where the base had prose atskills/larv-review/SKILL.md:58"Count" - Navigation:
README.md:70"Tools": names larv-review and metric-bench and the rules each holds - Locality:
bin/larv_review/main.py:238record_scoring: its result depends on which entry is newest in the history; the whole fall, scoring run, Ratchet, Still standing sequence must be held - Locality:
bin/larv_review/rules.py:71pending_falland:83standing_ratchet: the verdict depends on entry order in a file no single function shows - Locality, Shape:
bin/metric_bench/bench.py:7andbin/metric_bench/main.py:38: metric-bench depends on larv-review's history parser and keeps its owngitatbin/metric_bench/main.py:143 - Shape:
bin/larv_review/rules.py:16votesthrough:176scoring_run_due: one module named rules holds four concepts - Shape:
bin/larv_review/rules.py:35countedandbin/larv_review/history.py:68entries: one name over two things - Self-sufficiency:
bin/larv_review/seats.py:38reportandbin/larv_review/history.py:99heading: the seats file and history entry shapes are enforced in code
Inherited architect, decided by:
- Navigation, Self-sufficiency, Overall:
bin/larv_review/rules.py:125round_counts, withbin/larv_review/main.py:159round_count: the 3am question leads to one tested function, where the base hadskills/larv-review/SKILL.md:58"Count" - Navigation, Overall:
README.md:70"Tools": one hop from a symptom to a file - Locality:
bin/larv_review/main.py:238record_scoring: four helpers over four sources; whyfellis empty here and not atbin/larv_review/main.py:168is stated nowhere at the call - Locality:
bin/larv_review/rules.py:83standing_ratchetwithbin/larv_review/history.py:12RATCHET: state passes between commands through a markdown file's wording - Locality, Shape:
bin/larv_review/rules.py:176scoring_run_dueand:192emptied: the due rule packs five conditions - Shape:
bin/larv_review/rules.py:1the module: one name over many concepts - Shape:
bin/metric_bench/main.py:38andbin/metric_bench/bench.py:7: the history format has no home of its own;bin/larv_review/git.py:7HISTORYsits in a module named for git - Self-sufficiency:
bin/larv_review/main.py:1the module docstring: dense, but it ships with the code and states every rule
Maintainability senior, decided by:
- Navigation, Self-sufficiency, Overall:
bin/larv_review/rules.py:125round_counts, withbin/larv_review/main.py:159round_count: the rule is one small tested function, where the base hadskills/larv-review/SKILL.md:58"Count" - Navigation:
README.md:66"Tools": lists both tools with a pointer to each--help - Self-sufficiency:
bin/larv_review/rules.py:35counted,:50rulingand:60fallen: each rule stated in a docstring and enforced beside it, where the base hadskills/larv-review/SKILL.md:100"Ratchet" - Locality:
bin/larv_review/main.py:238record_scoring:due_reasonswith an empty fall, thenpending_fallseparately; four places to hold - Locality:
bin/larv_review/rules.py:71pending_falland:83standing_ratchet: order dependence between commands through a file - Locality, Shape:
bin/larv_review/seats.py:38report,bin/larv_review/history.py:99headingandbin/larv_review/main.py:267record_moved_mean: a one-time migration threaded through three modules - Shape:
bin/metric_bench/main.py:38andbin/metric_bench/bench.py:7: a second tool depends on the first one's parser;Entry.placesatbin/larv_review/history.py:40exists only for it - Shape:
bin/larv_review/rules.py:176scoring_run_due: a grab-bag module, offset by the truerskills/comprehension-panel/seat-prompts.md:1name - Self-sufficiency:
bin/larv_review/main.py:11theruling-baseparagraph: took three reads, resolved only bybin/larv_review/rules.py:134measured_commit
Junior A, decided by:
- Navigation:
bin/larv_review/rules.py:125round_counts: the symptom leads to theround-countcommand, where the base hadskills/larv-review/SKILL.md:58"Count" - Navigation:
README.md:68"Tools": names both tools;skills/larv-review/SKILL.md:46points at the same tool - Locality:
bin/larv_review/main.py:238record_scoring: whether the run blocks, ratchets or fails depends on three outside sources - Locality:
bin/larv_review/rules.py:134measured_commit: the base depends on what each past merged change added, by heading time - Locality:
bin/metric_bench/main.py:38entries, Entry: a second tool parses the history through larv-review's internals - Shape:
bin/larv_review/rules.py:176scoring_run_due: a catch-all module, offset by the clean split and the truerskills/comprehension-panel/seat-prompts.md:1 - Self-sufficiency:
skills/larv-review/SKILL.md:98"Ruling": the skill needsbin/larv_review/main.py:1open beside it, offset by rules now in code
2026-10-09T09:47:25Z, PR #232 at 324d650
Ratchet: the paired ruling counts as worse, as Locality fell and Locality is 6.50, under the floor of 7.0.
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6.5 | 6 | 6.5 |
| Mid A | 7 | 7 | 6.5 | 6.5 | 7 |
| Maintainability senior | 7.5 | 6.5 | 7 | 7 | 7 |
| Inherited architect | 7 | 6.5 | 6.5 | 7 | 7 |
| Mean | 7.13 | 6.50 | 6.63 | 6.63 | 6.88 |
Junior A, hardest first:
bin/larv_review/main.py:238record_scoring:due_reasonswith an empty fall means "reasons besides a fall", stated only atbin/larv_review/rules.py:93floor_verdictbin/larv_review/main.py:1the module docstring: 80 lines of interlocking rules that never say whybin/larv_review/rules.py:134measured_commit:not (last and last.fell)is true with no ruling at allbin/larv_review/rules.py:16votes, with:35countedand:40names_a_place: the counting rule spans three filesbin/larv_review/history.py:62entriesand:75seat_places: the parser and writer share a format by eyebin/loop_claude/cycle/judge.py:31run_judge: five outcomes, the spec a prompt stringbin/loop_claude/transcript_archive.py:30archive: reading-cost parses its outcome textbin/metric_bench/bench.py:87resolved: order-dependent legacy shim
Mid A, hardest first:
bin/larv_review/main.py:238record_scoring: six exits;due_reasonswith an empty fall list on purposebin/larv_review/rules.py:134measured_commit: states what, never whybin/larv_review/rules.py:176scoring_run_due: five inputs in one dense sentencebin/loop_claude/cycle/judge.py:31run_judge: two text checks against two textsbin/loop_claude/transcript_archive.py:30archive: a prose format that is a contract with another toolbin/loop_claude/cycle/states.py:69transition: case order decides behaviourbin/larv_review/rules.py:16votes: parse also enforces the ruling protocol;countedshadowedbin/loop_claude/claude_code/entries.py:45string_by_human: heuristics on who wrote an entry
Maintainability senior, hardest first:
bin/larv_review/main.py:238record_scoring: ten lines, six ideas;due_reasons's third argument means different things to its two callersbin/larv_review/rules.py:134measured_commit: the no-ruling case deduced only from the helpbin/larv_review/rules.py:176scoring_run_due: a 70-word docstring and a second document openbin/larv_review/rules.py:16votes: Overall's rule hidden inDIMENSIONS[-1];Nonemeans splitbin/larv_review/history.py:62entries: eight positional arguments counted against a regexbin/larv_review/history.py:99heading: a one-off migration threaded through three filesbin/metric_bench/bench.py:87resolved: depends on earlier bullets' orderbin/loop_claude/cycle/states.py:69transition: case order, resolved by comments
Inherited architect, would restructure first:
bin/loop_claude/transcript_archive.py:111finished_chunks,bin/reading_cost/archive.py:10project_archiveandbin/loop_claude/claude_code/entries.py:5parse_entry: lift the shared formats out of loop_claudebin/larv_review/rules.py:125round_countsand:176scoring_run_due: split rules into rounds, ruling, base and scoringbin/loop_claude/cycle/judge_reply.py:12JUDGE_RULESandbin/loop_claude/cycle/worker.py:41worker_args: give the prompt text one homebin/failure.py:5Failure,bin/file_writes.py:21replace_fileandbin/markdown_sections.py:6sections: tidy the bin rootbin/larv_review/git.py:10git,bin/metric_bench/main.py:143gitandbin/reading_cost/main.py:113located: one git runnerbin/loop_claude/project/checkout.py:32read_utf8: a generic reader under a lying namebin/larv_review/main.py:101main: an if/elif dispatch where the next command lands
2026-10-09T12:31:47Z, PR #234 at b1913ed, against 2157f59
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | same | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Navigation:
bin/transcript_archive/outcomes.py:4FINISHED_ATandfinished: one module now writes and parses the outcome lines, whichbin/loop_claude/cycle/run.py:83built inline and another file parsed - Locality, Navigation:
bin/reading_cost/main.py:35imports: reading-cost depends only on the sharedtranscript_archive, no longer on loop-claude's internals - Locality:
bin/loop_claude/paths.py:15paths_from: no longer reads the environment;LOOP_CLAUDE_TRANSCRIPTShas one home inbin/transcript_archive/location.py:5project_archive - Navigation, Self-sufficiency:
README.md:77"Tools": states the import root, the shared packages and where tests live - Self-sufficiency:
bin/transcript_archive/write.py:13archive: states the archive's layout at the write - Shape:
bin/loop_claude/loop/main.py:86run_loop: importsproject_key, which names the session and the models entry, from the archive package;bin/loop_claude/cycle/worker.py:7takesquotedfrom the archive's outcomes - Shape:
bin/loop_claude/cycle/judge.py:15JUDGE_RULES: judge_reply merged in, 183 lines
Inherited architect, decided by:
- Navigation, Shape:
bin/reading_cost/main.py:35main: imports one shared package named for what it holds, in place of loop-claude's private packages - Locality, Self-sufficiency:
bin/transcript_archive/outcomes.py:10finished: the outcome texts sit besideFINISHED_AT, the regex that reads them back - Shape:
bin/claude_code/session.py:21session_named: the Claude Code package depends on nothing in loop-claude - Locality:
bin/loop_claude/cycle/judge.py:15JUDGE_RULESand:93verdict_from: in one file withrun_judge - Locality:
bin/loop_claude/cycle/states.py:75transition: the cap no longer depends on case order - Navigation, Self-sufficiency:
README.md:77"Tools": states the layout and puts each test beside its package - Self-sufficiency:
bin/loop_claude/cycle/run.py:34Cycle.unfinished_evidence: the name says whatremainingneeded a comment for - Shape:
bin/loop_claude/loop/main.py:86run_loop,bin/loop_claude/cycle/worker.py:7worker_argsandbin/host_env.py:5host_env: small seams against the decoupling
2026-10-09T12:32:00Z, PR #234 at b1913ed
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6.5 | 6 | 6.5 |
| Mid A | 7 | 6 | 6.5 | 6 | 6.5 |
| Maintainability senior | 7 | 6.5 | 7 | 6.5 | 7 |
| Inherited architect | 7 | 6.5 | 6 | 6.5 | 6.5 |
| Mean | 7.00 | 6.25 | 6.50 | 6.25 | 6.63 |
Junior A, hardest first:
skills/larv-review/SKILL.md:55"Round cap": the counting rule spread across Round cap, Re-entry, Fresh and Suspendedbin/larv_review/main.py:197record_ruling: split, fell, refuse and stand branches;rulingreturns None for a split andfallenreuses itbin/larv_review/scoring.py:54scoring_run_due: why the next-release fallback holds is stated nowherebin/larv_review/base.py:14measured_commit: mechanics without purposebin/loop_claude/cycle/judge.py:126run_judge: the prompt assembled fromrecords.pyconstants; two texts, two matchersbin/reading_cost/measure.py:21Checkout:rootandtopread as synonyms
Mid A, hardest first:
bin/larv_review/main.py:197record_ruling: said and counted, None as split, three fall stepsbin/comprehension_history.py:74entries: the history format is defined only by its writerbin/loop_claude/cycle/judge.py:126run_judge: elided evidence fails, a misquoted answer becomes an interruptbin/larv_review/scoring.py:54scoring_run_due: one dense sentence over two plans; the module also parses the planbin/larv_review/base.py:14measured_commit: the why lives in the help and the skillbin/loop_claude/cycle/states.py:69transition:quoted_answersthreaded through three states and two files
Maintainability senior, hardest first:
bin/larv_review/main.py:197record_ruling: said and count, ruled reassigned, a silent fall-through;ruling's None means two thingsbin/larv_review/base.py:14measured_commit: a three-step fallback across three files, ordered by an ISO string unsaidbin/comprehension_history.py:74entries: the format is never stated in one placebin/loop_claude/cycle/judge.py:15JUDGE_RULES: the prompt interpolates four constants from another modulebin/larv_review/scoring.py:54scoring_run_due: five inputs;scoringalso parses the planbin/rank_todo.py:416rank_details: regex surgery in a script larv-review imports as a library
Inherited architect, would restructure first:
bin/loop_claude/project/records.py:6QUOTE_SOURCES: prompt fragments spliced into the judge, worker and interrupt textsbin/transcript_archive/location.py:10project_key: the project's identity filed under the archivebin/host_env.py:5host_env: a test helper among the production modulesbin/larv_review/rounds.py:8round_counts: only counts; the cap lives in the skillbin/larv_review/base.py:1the module: the ruling base, named like a base classbin/larv_review/scoring.py:6imports: one tool imports another tool's internals frombin/rank_todo.pybin/loop_claude/loop/main.py:1the docstring: restates constants it cannot keep true
2026-10-09T12:46:51Z, PR #234 at dde06fc, against 2157f59
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Navigation, Locality, Shape:
bin/transcript_archive/location.py:7project_archive: the archive's location lives in one shared package, where reading-cost reached into loop-claude's internals - Locality, Shape, Self-sufficiency:
bin/loop_claude/cycle/run.py:83worker_start:worker_argsbuilds only flags and each prompt is a named function inbin/loop_claude/cycle/prompts.py:41 - Locality, Self-sufficiency:
bin/transcript_archive/outcomes.py:10finished: the outcomes format's writer and reader sit in one file - Shape:
bin/larv_review/ruling.py:42split:rulingno longer returns None for a split - Shape, Locality:
bin/larv_review/main.py:204record_ruling: two plain checks in place offall_step's strings; plan logic inbin/larv_review/plan.py:5 - Navigation:
README.md:77"Tools": states the import root and the shared packages - Self-sufficiency:
bin/comprehension_history.py:74entries: a worked example of an entry above the parser - Self-sufficiency, Shape:
bin/metric_bench/bench.py:88tracked_places: the rename garbled the docstring - Locality:
bin/loop_claude/cycle/prompts.py:60quoted: duplicatesbin/transcript_archive/outcomes.py:31quoted
Inherited architect, decided by:
- Navigation, Locality:
bin/transcript_archive/outcomes.py:10finished: every outcome line andFINISHED_ATsit in one module - Navigation, Locality:
bin/reading_cost/main.py:35imports: reading-cost no longer reaches into loop-claude's internals - Navigation, Self-sufficiency:
README.md:77"Tools": the layout is stated - Shape, Locality:
bin/todo_table.py:27items: the table format is its own module, not a CLI script used as a library - Shape:
bin/larv_review/plan.py:5scoring_run_due: the plan trigger with its why, apart from the floor arithmetic - Locality, Shape:
bin/larv_review/ruling.py:42split: split, ruling and lead are separate - Locality, Shape:
bin/loop_claude/cycle/prompts.py:41fresh: every prompt text in one module - Shape:
bin/comprehension_history.py:67ScoringRun: one name no longer covers two concepts - Self-sufficiency:
bin/larv_review/main.py:5the help: defines the ratchet, the paired ruling and the scoring run first - Locality:
bin/loop_claude/cycle/states.py:75transition: the past-the-cap case has its own guard - Shape:
bin/metric_bench/bench.py:87tracked_places: the docstring reads worse after the rename
2026-10-09T12:46:51Z, PR #234 at dde06fc
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 6 | 6 | 6 |
| Mid A | 7.5 | 6.5 | 7 | 6.5 | 7 |
| Maintainability senior | 7 | 6.5 | 7 | 7 | 7 |
| Inherited architect | 7 | 6.5 | 6.5 | 7 | 7 |
| Mean | 7.13 | 6.50 | 6.63 | 6.63 | 6.75 |
Junior A, hardest first:
bin/larv_review/main.py:204record_ruling: six outcomes, said against count,[-1]meaning Overallbin/larv_review/ruling.py:8votes: a parser that also enforces the seat-count rulebin/larv_review/ruling_base.py:14measured_commit: the docstring explains only after the code confusedbin/larv_review/plan.py:5scoring_run_due: four ideas at once;emptiedtraced by handbin/metric_bench/bench.py:87tracked_places: a garbled docstringbin/loop_claude/cycle/states.py:69transition: the quoted-answer cap across three files
Mid A, hardest first:
skills/larv-review/SKILL.md:55"Round cap": four subsections pointing at each otherbin/larv_review/main.py:204record_ruling: five outcomes; the duplicate check in two placesbin/loop_claude/cycle/judge.py:95run_judge: many endings, two quote checksbin/loop_claude/cycle/states.py:69transition: the cap counter rides through three statesbin/larv_review/ruling_base.py:14measured_commit: needs the history format and first-parent semanticsbin/reading_cost/measure.py:27measure: string-compared timestamps and approximate shell parsing
Maintainability senior, hardest first:
bin/larv_review/main.py:204record_ruling: five outcomes, competing local namesskills/larv-review/SKILL.md:55"Round cap": a chain of cross-referencesbin/comprehension_history.py:103seat_places: depends unsaid onsectionsdropping blank linesbin/larv_review/plan.py:5scoring_run_due: four inputs; the rule restated in the helpbin/larv_review/ruling_base.py:14measured_commit: why same is safe needs the skillbin/metric_bench/bench.py:87tracked_places: a garbled docstring
Inherited architect, would restructure first:
bin/larv_review/rounds.py:12round_counts: the round cap lives in prose; a mis-cased title escapes the countbin/loop_claude/cycle/judge.py:17Reply: the judge contract spread over four filesbin/loop_claude/project/checkout.py:28read_utf8: a generic reader under the checkout's namebin/test_env.py:5isolated_env: test support named like a testbin/larv_review/git.py:6imports: the git adapter depends on a domain typebin/loop_claude/cycle/prompts.py:60quoted: two copies, withSTAMPand the git runner also duplicated
2026-10-09T12:53:41Z, PR #234 at b346a35, against 2157f59
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Navigation, Locality:
bin/reading_cost/main.py:35main: the archive reader and location come from a shared package, not loop-claude's internals - Locality, Shape:
bin/loop_claude/cycle/run.py:83worker_start: picks one of three named prompts for a four-argumentworker_args - Navigation, Shape:
bin/loop_claude/cycle/prompts.py:1the module: the texts the worker and the human see sit together - Locality, Shape:
bin/larv_review/main.py:204record_ruling: plain checks withsplit,leadandoverall - Navigation, Shape:
bin/larv_review/ruling_base.py:4measured_commit: a truthful module name;HistoryChangebeside the format - Navigation, Locality:
bin/todo_table.py:27items: the table parser is shared, not reached into a CLI script - Navigation, Self-sufficiency:
README.md:77"Tools": states the layout - Self-sufficiency:
bin/comprehension_history.py:82entries: a worked example above the parser - Shape:
bin/reading_cost/measure.py:20Checkout: fields say which root keys the archive - Locality:
bin/loop_claude/cycle/states.py:75transition: the past-the-cap case has its own guard
Inherited architect, decided by:
- Locality, Shape:
bin/transcript_archive/location.py:7project_archive: both tools import one shared package - Navigation, Locality:
bin/transcript_archive/outcomes.py:10finished: the outcome line sits beside the regex that reads it back - Locality, Shape:
bin/loop_claude/cycle/run.py:83worker_start:worker_argstakes the finished prompt - Shape:
bin/larv_review/git.py:5imports: the git helper no longer depends on the base search - Shape, Self-sufficiency:
bin/larv_review/plan.py:5scoring_run_due: its own module, saying why each trigger exists - Self-sufficiency, Locality:
bin/larv_review/ruling.py:27overalland:47split: no None meaning split, no bare index - Navigation:
bin/reading_cost/measure.py:20Checkout: the fields name which root - Self-sufficiency:
bin/comprehension_history.py:82entries: a worked example - Navigation:
README.md:77"Tools": the import root and the tests' place - Shape:
bin/larv_review/main.py:234recorded_ruling: eight positional parameters, a step back - Self-sufficiency:
bin/loop_claude/cycle/prompts.py:29quoted: a second copy
2026-10-09T12:53:41Z, PR #234 at b346a35
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 7 | 6 | 6.5 |
| Mid A | 7.5 | 7 | 7 | 6.5 | 7 |
| Maintainability senior | 7.5 | 7 | 7 | 7 | 7 |
| Inherited architect | 7.5 | 7 | 6.5 | 7 | 7 |
| Mean | 7.38 | 6.75 | 6.88 | 6.63 | 6.88 |
Junior A, hardest first:
skills/larv-review/SKILL.md:55"Round cap": rules pointing at each other; Fresh mixes five thingsbin/larv_review/main.py:204record_ruling: five exits; the counting rule inbin/larv_review/ruling.py:32countedbin/larv_review/ruling_base.py:4measured_commit: a set difference of headings per changebin/loop_claude/cycle/judge.py:16JUDGE_RULES: a prompt built from constants in three files; two quote checksbin/comprehension_history.py:95entries: two label regexes tied to the writer's wordingbin/reading_cost/shell_reads.py:52segments: heuristic shell parsing
Mid A, hardest first:
bin/larv_review/main.py:204record_ruling: five branches across four modulesbin/larv_review/ruling_base.py:4measured_commit: reread three times to trust itbin/loop_claude/cycle/judge.py:127run_judge: two quote checks with opposite policiesbin/loop_claude/cycle/states.py:69transition: the quoted-answer counter through three statesbin/comprehension_history.py:95entries: Markdown parsed back into recordsbin/metric_bench/bench.py:87tracked_places: order-dependent legacy resolution
Maintainability senior, hardest first:
bin/larv_review/main.py:204record_ruling: six exits;fresh=Falsehides a skipped checkbin/larv_review/ruling_base.py:4measured_commit: why needs the skillbin/loop_claude/cycle/judge.py:16JUDGE_RULES: assembled across three filesbin/comprehension_history.py:95entries: the format defined by example and regexesbin/loop_claude/cycle/states.py:69transition: the counter survives an interruptbin/reading_cost/shell_reads.py:52segments: opaque on what it misses
Inherited architect, would restructure first:
bin/larv_review/rounds.py:8round_counts: the cap lives in prose; the tool only reportsbin/larv_review/main.py:204record_ruling: the record commands in a 326-line mainbin/larv_review/git.py:8git: four git wrappers, one archive key computed two waysbin/loop_claude/cycle/prompts.py:29quoted: duplicated, as isSTAMPbin/larv_review/ruling_base.py:24base: a name that reads as a base class
2026-10-09T13:04:47Z, PR #234 at b2a277b, against 2157f59
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Navigation:
bin/transcript_archive/location.py:7project_archive: the archive's writer and readers share one package; outcome strings sit beside the regex that reads them - Locality, Shape:
bin/loop_claude/cycle/run.py:83worker_start: picks the prompt in three lines;worker_argsonly builds argv - Shape, Navigation:
bin/larv_review/main.py:100main: only dispatch; recording and plan reading have their own modules - Locality, Navigation:
bin/todo_table.py:27items: table parsing shared, not reached into a tool script - Shape:
bin/larv_review/ruling.py:47split: split, lead and overall each say what they compute - Shape, Self-sufficiency:
bin/reading_cost/measure.py:20Checkout: fields name which root - Locality:
bin/claude_code/session.py:21session_named: the adapter takes only a name and a folder - Navigation:
README.md:77"Tools": names the import root and where tests live - Self-sufficiency:
bin/larv_review/main.py:5the help: explains the ruling, ratchet, scoring run and floor first
Inherited architect, decided by:
- Locality, Navigation:
bin/transcript_archive/location.py:7project_archive: location, writer and reader are peers both tools import - Locality:
bin/transcript_archive/outcomes.py:6FINISHED: the format and its parser in one module - Shape, Navigation:
bin/larv_review/plan.py:8scoring_run_due: the plan trigger has its own file;scoring.pyholds only the floor - Shape:
bin/larv_review/record.py:14record_ruling: recording is one module, its guards inline - Locality, Self-sufficiency:
bin/loop_claude/cycle/worker.py:40worker_args: takes a ready prompt; the texts are named functions - Shape:
bin/loop_claude/cycle/judge.py:16JUDGE_RULES: the judge is one file - Locality:
bin/loop_claude/paths.py:15paths_from: the env override is read in one place - Self-sufficiency:
bin/comprehension_history.py:81entries: a worked example above the parser - Navigation, Self-sufficiency:
README.md:77"Tools": the layout is stated and checked - Locality:
bin/larv_review/ruling.py:52ruling: returns same for a split, so onlysplitguards it
2026-10-09T13:04:48Z, PR #234 at b2a277b
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 7 | 6 | 6.5 |
| Mid A | 7.5 | 6.5 | 7 | 6.5 | 6.5 |
| Maintainability senior | 7.5 | 6.5 | 7 | 6.5 | 7 |
| Inherited architect | 7.5 | 7 | 6.5 | 7 | 7 |
| Mean | 7.38 | 6.50 | 6.88 | 6.50 | 6.75 |
Junior A, hardest first:
bin/larv_review/record.py:14record_ruling: seven exits; the terms only in the help and the skillbin/larv_review/ruling_base.py:4measured_commit: diffs heading sets per changebin/comprehension_history.py:94entries: regex groups used by positionbin/larv_review/ruling.py:32counted: the counting rule stated in the help, not herebin/loop_claude/cycle/states.py:69transition: the cap counter across three filesbin/loop_claude/cycle/judge.py:127run_judge: two quote checks with different rules
Mid A, hardest first:
skills/larv-review/SKILL.md:82"Architecture": conditions with exceptions; Fresh took three rereadsbin/larv_review/record.py:14record_ruling: six outcomes across five modulesbin/loop_claude/cycle/judge.py:16JUDGE_RULES: one sentence takes three filesbin/larv_review/ruling_base.py:4measured_commit: the fallback chain over two filesbin/larv_review/plan.py:8scoring_run_due: release sets held at oncebin/comprehension_history.py:94entries: the format pinned in three places
Maintainability senior, hardest first:
bin/larv_review/record.py:14record_ruling: five exits; the rule spread over five modulesbin/loop_claude/cycle/judge.py:16JUDGE_RULES: instructions stitched from three modulesbin/larv_review/plan.py:8scoring_run_due: five inputs; the docstring restates the helpbin/larv_review/ruling_base.py:4measured_commit: opaque without its docstringbin/larv_review/seats.py:17SeatsFile:readis a noun; locals rename one pair two waysbin/comprehension_history.py:110seat_places: a line-state parser two tools depend on
Inherited architect, would restructure first:
bin/larv_review/rounds.py:4PASSES: the cap is the skill's; the pass list has two homesbin/comprehension_history.py:149ruling_entry: reader, writer and seat registry in one root modulebin/larv_review/record.py:67read_fall_scores: a parameter named votes holds a headingbin/larv_review/git.py:8git: four ways to run git; the archive key's root differs between toolsbin/larv_review/record.py:14record_ruling: branches on worse, fell, scores file and split at once
2026-10-09T13:23:40Z, PR #234 at 7e13778, against 2157f59
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Navigation, Shape:
bin/larv_review/record.py:14record_ruling: recording, dispatch and plan reading each in a file named for it - Shape, Navigation:
bin/larv_review/plan.py:8scoring_run_due: file names tell the truth; todo parsing inbin/todo_table.py:27items - Navigation, Shape:
bin/transcript_archive/location.py:7project_archive: both tools share top-level packages, not another tool's internals - Locality:
bin/loop_claude/cycle/run.py:83worker_start: a prompt changes without reading the argument plumbing - Locality:
bin/loop_claude/cycle/states.py:75transition: the past-the-cap case has its own guard - Locality:
bin/transcript_archive/location.py:7project_archive: the env override resolved in one place - Self-sufficiency:
bin/larv_review/main.py:1the help: defines seats, the ruling, the ratchet and the floor first - Shape, Self-sufficiency:
bin/reading_cost/measure.py:20Checkout: fields say which root keys the archive
Inherited architect, decided by:
- Locality, Navigation:
bin/transcript_archive/outcomes.py:6FINISHED: outcome lines written and parsed in one file - Locality, Shape:
bin/transcript_archive/location.py:7project_archive: reading-cost no longer reaches into loop-claude - Navigation, Locality:
bin/loop_claude/cycle/judge.py:94verdict_from: the 3am question lands in one file - Shape, Navigation:
bin/larv_review/main.py:102main: dispatch only - Locality, Shape:
bin/todo_table.py:27items: larv-review imports no tool's internals - Locality:
bin/larv_review/record.py:32record_ruling: the fall handling reads in place - Locality, Shape:
bin/loop_claude/cycle/run.py:83worker_start: the start chosen at the call site - Shape:
bin/reading_cost/measure.py:20Checkout: names that say which root - Self-sufficiency:
bin/comprehension_history.py:81entries: a sample entry above the parser - Locality:
bin/loop_claude/cycle/states.py:75transition: no case-order dependence
2026-10-09T13:23:40Z, PR #234 at 7e13778
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7.5 | 7 | 7 | 6.5 | 7 |
| Mid A | 7 | 6.5 | 7 | 6.5 | 6.5 |
| Maintainability senior | 7 | 6.5 | 7 | 6 | 6.5 |
| Inherited architect | 7 | 6.5 | 6 | 7 | 6.5 |
| Mean | 7.13 | 6.63 | 6.75 | 6.50 | 6.63 |
Junior A, hardest first:
bin/larv_review/record.py:14record_ruling: four outcomes, two parallel vote dictsbin/larv_review/ruling.py:32counted: an asymmetric rule;PLACEin another filebin/metric_bench/python_units.py:86score: needs a paper outside the repobin/comprehension_history.py:94entries: leans onsectionsdropping blank linesbin/loop_claude/cycle/judge.py:127run_judge: two quote checks against two textsskills/larv-review/SKILL.md:55"Round cap": sections pointing at each other
Mid A, hardest first:
bin/larv_review/record.py:14record_ruling: five outcomes;readas a nounbin/comprehension_history.py:110seat_places: depends unsaid onsectionsdropping blank linesbin/larv_review/ruling_base.py:4measured_commit: several ideas held at oncebin/loop_claude/cycle/judge.py:16JUDGE_RULES: decision logic in Englishbin/claude_code/entries.py:28written_by_human: heuristics over the transcript shapebin/loop_claude/loop/reload.py:50exec_if_edited: any tool's edit reloads the loop
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:82"Architecture": six interlocking conditionsbin/larv_review/record.py:14record_ruling: two checks run twice; the rule whole only in the helpbin/comprehension_history.py:94entries: a format shared by three tools by string conventionbin/larv_review/plan.py:8scoring_run_due: three things combined in "work has landed"bin/loop_claude/cycle/judge.py:16JUDGE_RULES: behaviour in prose, post-filtered three waysbin/rank_todo.py:357rank_details: rebuilds the file by slicing
Inherited architect, would restructure first:
bin/larv_review/rounds.py:8round_counts: the cap is prose in the skillbin/larv_review/record.py:14record_ruling: an if-chain with prints interleavedbin/larv_review/scoring.py:8fall_ruling: a ruling rule in the scoring modulebin/claude_code/trust.py:10trust: effects in a package documented as shapesbin/larv_review/git.py:8git: git run five ways
2026-10-12T21:36:23Z, PR #241 at 7cf4f54, against 4cc9953
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | same | same |
| Inherited architect | same | worse | same | same | same |
Mid A, decided by:
- Shape:
bin/rule_hook/rules.py:140file_breaches: four rules in one function, picked by file-name tests, each a comprehension of three or four conditions - Shape:
bin/rule_hook/rules.py:86git_commit: a hand-written git argument parser with two index-juggling loops and a nested loop that movesat - Shape:
bin/rule_hook/main.py:1module docstring: it andRULESeach describe the six rules in their own words, as do CLAUDE.md, CHANGELOG.md and README.md - Shape:
bin/rule_hook/main.py:93checked_call: returns a 3-tuple or None, and its name hides that it replays a MultiEdit's edits
Inherited architect, decided by:
- Locality:
skills/larv-review/SKILL.md:58"Opt-out": an opted-out chunk skips a scoring runlarv-review needs-scoring-runstill calls due, and "End" adds a third kind of run - Locality:
bin/rule_hook/main.py:76main: logs each granted override before it knows whether another rule still blocks the call, into an unversioned log keyed by branch name and never pruned - Locality:
bin/rule_hook/main.py:115todo_breaches: the hook importsrank_todo.check_failures, so editing rank_todo's messages changes what blocks a commit, while README calls rank_todo.py a tool of one file; the rule's logic sits in main.py, apart from the other rules in rules.py
2026-10-12T22:10:36Z, PR #241 at 7598584, against 4cc9953
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | same | same |
| Inherited architect | same | worse | same | same | same |
Mid A, decided by:
- Shape:
bin/rule_hook/rules.py:8RULES: the module holds the rule table, override parsing and git command-line parsing, whilegit.pyonly runs git - Shape:
bin/rule_hook/override_log.py:82logged: builds an Entry though it reads as a yes/no check;holdsis an ancestry test andwhatsays little - Shape:
bin/rule_hook/main.py:66main: passesverdicttwo bare booleans - Locality:
bin/shell_command.py:20simple_commands: shorter reading-cost parsing, against a parser change for the hook moving reading-cost figures; the two cancel - Locality:
bin/rule_hook/main.py:99bash_breaches: parses the command's commits twice
Inherited architect, decided by:
- Locality:
bin/shell_command.py:31segments: one grown parser serves reading-cost and the hook, so tightening it for one moves the other - Locality:
bin/rule_hook/override_log.py:36this_chunks: chunk membership rests on reflog expiry and a default-branch guess, with three failures in one nested ternary at line 96 - Locality:
bin/rule_hook/main.py:82answer: resolve-then-write order carried by one comment, andBreach.directoryset deep in rules.py decides which log gets a line - Locality:
bin/rule_hook/rules.py:108git_subcommand: a second answer to "is this a git commit" besidemeasure.py'sCOMMITregex - Locality:
bin/rule_hook/todo_check.py:7import rank_todo: a change to rank_todo's check changes what the hook blocks
2026-10-12T22:43:27Z, PR #242 at 63b0759, against 4cc9953
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | better | better | better |
| Inherited architect | better | same | same | same | better |
Mid A, decided by:
- Shape, Navigation:
bin/shell_command.py:20simple_commands: splitting a shell command and tracking eachcdis one module named for what it holds, whereshell_reads.pyspread it over three helpers - Shape:
bin/reading_cost/measure.py:7measureimports:resolvedcomes fromshell_command, no longer from a module named for shell reads - Self-sufficiency, Navigation:
bin/rule_hook/main.py:1module docstring: each rule'sholdsandcheckedbuild--help, so a blocked commit leads to one authority - Navigation:
CLAUDE.md:50"Topic rules": each topic's rules have their own file and the section says which loads with what, where the old charter was one 167-line wall - Shape:
bin/rule_hook/rules.py:108git_subcommand: git argument parsing sits inrules, whilegit.pyonly runs git - Locality:
bin/loop_claude/project/records.py:14read_records: the rules folder is spelled twice, asHUMAN_RULES_DIRand asclaude_md.parent / "rules" - Shape:
bin/rule_hook/todo_check.py:7import rank_todo: one tool imports another tool's script, though the README states it
Inherited architect, decided by:
- Navigation, Overall:
CLAUDE.md:50"Topic rules": each topic has a file named for it and scoped bypaths:, indexed in four lines, so a rule is one step from its question - Navigation:
bin/shell_command.py:20simple_commands: command splitting has one obvious home, used by the hook and by reading-cost - Locality:
bin/shell_command.py:31segments: the shared parser means a change for the hook moves reading-cost's figures, and nothing atreading_cost/says so - Shape:
bin/rule_hook/rules.py:108git_subcommand: a module namedrulesparses git's command line, andBreach.directorycarries the override log's concern - Shape, Locality:
bin/rule_hook/override_log.py:96print_overrides: a chained ternary of four outcomes besidethis_chunks, whose reflog rule needs both comments held at once - Locality:
bin/rule_hook/main.py:82answer: resolve-before-write holds only by the loops' arrangement, stated in one comment - Self-sufficiency:
bin/rule_hook/main.py:1module docstring: the rules print fromRULES, whileCLAUDE.md:67now sends a charter reader torules.pyfor the pinning rule - Locality:
bin/rule_hook/todo_check.py:10todo_breaches: a change torank_todo's output reaches a hook that can block every commit
2026-10-12T22:56:19Z, PR #242 at f351e14, against 4cc9953
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | same | same |
| Inherited architect | same | same | same | same | same |
Mid A, decided by:
- Shape:
bin/rule_hook/rules.py:99commits: a file named for the rule table parses git's command line, whilegit.pyonly runs git - Shape:
bin/rule_hook/override_log.py:82logged: builds an Entry and logs nothing;holdsis an is-ancestor test andthis_chunksneeds two comments - Shape:
bin/shell_command.py:20simple_commands: shell lexing moved out ofshell_reads.py, which now holds only what counts as a read
Inherited architect, decided by:
- Locality, Shape:
bin/shell_command.py:20simple_commands: one shared lexer, though a fix for the hook shifts reading-cost's figures unseen - Locality, Shape:
bin/rule_hook/rules.py:108git_subcommand: a second definition of a commit besidemeasure.py'sCOMMITregex, in a module that also holds the rule table and verdict - Locality:
bin/rule_hook/override_log.py:86print_overrides: a four-way chained conditional besidethis_chunks' reflog rule - Locality:
skills/larv-review/SKILL.md:56"Opt-out": a third skip mechanism to cross-check beside "Precedent, then ask" and "Suspended" - Navigation, Self-sufficiency:
bin/rule_hook/main.py:46main: a blocked call leads cold to the rule, its text and the override syntax - Self-sufficiency, Navigation:
CLAUDE.md:50"Topic rules": each topic's rules are found by file name, whileCLAUDE.md:67sends a reader torules.pyfor the pinning rule's full text
2026-10-12T23:45:13Z, PR #245 at b568522, against 4cc9953
Ruling: worse
Fell: Navigation, Locality. Under the floor of 7.0: Locality 6.88, so the same ruling counts as worse.
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | same | same |
| Inherited architect | worse | worse | same | same | worse |
| Maintainability senior | worse | worse | same | same | worse |
| Junior A | worse | same | same | same | same |
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 6.5 | 6.5 | 6.5 |
| Mid A | 7.5 | 7 | 7 | 6.5 | 7 |
| Maintainability senior | 7.5 | 7 | 7 | 7 | 7 |
| Inherited architect | 7.5 | 7 | 6.5 | 7 | 7 |
| Mean | 7.38 | 6.88 | 6.75 | 6.75 | 6.88 |
Mid A, decided by:
- Locality:
agents/prose-reviewer.md:9"The bar": the prose bar now combines CLAUDE.md → "Prose", the standards files and path-matched rules/ files, andprose.mdsplits one concept across two files - Locality:
bin/loop_claude/project/records.py:11HUMAN_RULE_NAMES: what the judge reads depends on the files in two folders, andQUOTE_SOURCESslices the tuple to word it - Overall, Shape:
bin/shell_command.py:20simple_commands:shell_readsdrops its own lexer for a shared module, at the cost of denser regexes - Shape:
bin/rule_hook/rules.py:99commits: git argument parsing sits in a module named for the rule registry - Self-sufficiency:
bin/rule_hook/main.py:1module docstring: each rule carries its own text and --help is generated from it - Navigation:
CLAUDE.md:36"Topic rules": a router to rules/, and the standards are a third home for instructions, balanced by README's rule-hook entry
Inherited architect, decided by:
- Navigation, Locality:
CLAUDE.md:21"Prose": each priority is split between CLAUDE.md andskills/larv-review/standards/prose.md, and nothing in CLAUDE.md points to the second - Navigation:
agents/prose-reviewer.md:9"The bar": the bar is a three-way split, the worst place to go from a flag to the rule that fired - Locality:
bin/loop_claude/project/records.py:10HUMAN_RULE_DIRS: the judge globs one skill's internal folder, and moving it silently empties the records - Locality:
bin/loop_claude/cycle/prompts.py:5QUOTE_SOURCES: a tuple sliced with [:-1] and [-1] to place an "or", worked out across two modules - Locality, Shape:
bin/shell_command.py:20simple_commands: lexing moved out ofshell_readsinto a shared module, the change's clearest gain - Shape:
bin/rule_hook/rules.py:108git_subcommand: git command parsing sits in rules.py while git.py only runs git - Navigation:
README.md:75"Tools": the new tool is easy to find cold, and --help renders each rule - Locality:
bin/rule_hook/override_log.py:96print_overrides: a three-level conditional expression, and "this chunk" needs reflog semantics
Maintainability senior, decided by:
- Navigation, Locality:
skills/larv-review/standards/prose.md:3"Prose, in detail": it expands CLAUDE.md's priorities by number, so it reads only with CLAUDE.md open, and a change to a priority edits two files - Navigation:
CLAUDE.md:36"Topic rules": the router lists only rules/, so finding where comments or "A false claim is a bug" live needs README line 3 or the Vocabulary - Locality:
skills/larv-review/standards/quality.md:7"Quality": each standard points at two or three other files, and CLAUDE.md:50 paraphrases the README rule without pointing to it - Locality:
bin/loop_claude/project/records.py:9HUMAN_RULE_NAMES: a derived tuple sliced by position feeds three prompts, and the folder list must match hand-kept copies in CLAUDE.md and AGENTS.md - Locality:
bin/shell_command.py:31segments: reading-cost and rule-hook share one parser, so a fix for the hook moves the reading-cost figures - Navigation:
bin/rule_hook/main.py:1module docstring: from a blocked commit to the rule is easy through --help andRULES
Junior A, decided by:
- Navigation:
CLAUDE.md:36"Topic rules": the moved rules are not routed from here, and nothing in CLAUDE.md points tostandards/quality.md - Navigation:
skills/larv-review/standards/prose.md:3"Prose, in detail": one rule lives in two files, and a reader must know the second exists - Navigation:
agents/prose-reviewer.md:9"The bar": working out which documents bind a review takes a small search
Junior A, hardest first:
skills/larv-review/SKILL.md:64"Round cap": with "Re-entry" and "Fresh", four sections held at once to count a roundbin/larv_review/record.py:14record_ruling: four exits, its terms defined in three other filesbin/rule_hook/override_log.py:36this_chunks: reflog semantics, and a three-way conditional at :96bin/shell_command.py:31segments: dense regexes and a redirect flag, output unpredictable by readingbin/loop_claude/cycle/judge.py:16JUDGE_RULES: the judge's prompt assembled from three filesbin/larv_review/plan.py:8scoring_run_due: set arithmetic over four inputs
Mid A, hardest first:
bin/larv_review/record.py:14record_ruling: four outcomes, the counting rule only in the helpskills/larv-review/SKILL.md:64"Round cap": interlocking with "Re-entry" and "Fresh"bin/loop_claude/cycle/judge.py:127run_judge: two mismatches handled two ways;QUOTE_SOURCESslices a tuplebin/loop_claude/cycle/states.py:69transition: a counter threaded through three statesbin/rule_hook/override_log.py:86print_overrides: three chained conditionalsbin/larv_review/ruling_base.py:4measured_commit: rests on the docstring and the help
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:91"Architecture": six conditions held at once, and the no-rescore file list only in prosebin/larv_review/record.py:14record_ruling: five modules open at oncebin/larv_review/ruling_base.py:4measured_commit: why same does not move the base, on a second readbin/loop_claude/cycle/judge.py:16JUDGE_RULES: what the judge accepts spans three files nothing checks agreebin/rule_hook/override_log.py:36this_chunks: reflog membership; twodefault_branchfunctions differbin/rank_todo.py:374rank_details: regex surgery on markdown
Inherited architect, would restructure first:
bin/loop_claude/project/records.py:16read_records: a missing rule folder is skipped silentlybin/rule_hook/override_log.py:110default_branch: a second default-branch rule, unlike larv-review'sbin/loop_claude/loop/main.py:16module docstring: --help restates the retry constants as prosebin/larv_review/rounds.py:4PASSES: the round cap and rescores live in the skill's prosebin/comprehension_history.py:42Entry: one name over five concepts across the repo
2026-10-13T05:52:26Z, PR #245 at f6d5ea7, against 4cc9953
Ruling: worse
Fell: Shape. Under the floor of 7.0: Shape 6.75, so the same ruling counts as worse.
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | same | same |
| Inherited architect | same | worse | worse | same | worse |
| Maintainability senior | same | worse | same | same | same |
| Junior A | better | same | worse | same | same |
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 7 | 6 | 6.5 |
| Mid A | 7.5 | 6.5 | 7 | 6.5 | 7 |
| Maintainability senior | 7.5 | 7 | 6.5 | 6.5 | 7 |
| Inherited architect | 7 | 6.5 | 6.5 | 7.5 | 7 |
| Mean | 7.25 | 6.63 | 6.75 | 6.63 | 6.88 |
Mid A, decided by:
- Shape:
bin/rule_hook/rules.py:108git_subcommand: a module named for the rule catalogue also parses git arguments, override markers and verdicts - Shape:
bin/rule_hook/override_log.py:96print_overrides: three chained conditional expressions decide whether the reflog is readable - Shape:
bin/rule_hook/rules.py:73Verdict: two unrelatedVerdicttypes, andlogged,existingandholdsdo not say what they compute - Shape:
bin/loop_claude/project/records.py:11HUMAN_RULE_NAMES: a nested comprehension where one constant stood, thougheitherkeeps the prompt readable - Shape:
bin/shell_command.py:20simple_commands: a flatshell_readsover a denser shared lexer, a small net change
Inherited architect, decided by:
- Shape:
bin/rule_hook/rules.py:108git_subcommand: rules.py mixes six jobs while git.py only runs git, andreading_cost/measure.py:9COMMITis a second commit check - Locality:
CLAUDE.md:21"Prose": the worker's core andskills/larv-review/standards/prose.mdare two copies that word the contrastive tail differently - Locality:
bin/loop_claude/project/records.py:10HUMAN_RULE_DIRS: the judge hard-codes larv-review's folder, and a test imports the constant back out - Locality, Shape:
bin/rule_hook/override_log.py:86print_overrides: a three-deep conditional, reflog semantics in a docstring, and the git-dir lookup done again - Shape:
bin/rule_hook/main.py:82answer: a two-phase pass and an unnamed 3-tuple frombreaches_of - Locality, Shape:
skills/larv-review/SKILL.md:56"Opt-out": a third skip mechanism, three sections held at once
Maintainability senior, decided by:
- Locality:
CLAUDE.md:21"Prose": the prose rules exist twice, andCLAUDE.md:30words the contrastive tail differently fromskills/larv-review/standards/prose.md:13 - Locality:
rules/decisions.md:9"Decisions": the decision rule is split between a worker file andskills/larv-review/standards/comments.md:9"Where a choice goes" - Locality:
bin/rule_hook/override_log.py:110default_branch: a second default-branch rule and git runner, unlike larv-review's - Locality:
bin/loop_claude/project/records.py:10HUMAN_RULE_DIRS: one folder list feeds the judge, the quote sources and a test, andloop/main.pyspells it out by hand - Locality:
bin/rank_todo.py:114check_failures: copies the read, rank and compare flowmainstill runs - Locality:
bin/rule_hook/rules.py:99commits: computed on two paths for one command
Junior A, decided by:
- Navigation:
CLAUDE.md:38"Topic rules": a short router to one small file per topic - Navigation:
README.md:75"Tools": names rule-hook and what runs it - Shape:
bin/rule_hook/rules.py:108git_subcommand: the rule table's file also parses git arguments, Dockerfiles and ignore files - Shape:
bin/rule_hook/override_log.py:96print_overrides: three failure cases in one chained conditional - Shape:
bin/rule_hook/main.py:38imports: the entry point pulls from six modules - Navigation:
bin/shell_command.py:20simple_commands: shell tokenizing has one home
Junior A, hardest first:
bin/larv_review/record.py:14record_ruling: five files held at once, four outcomesbin/rule_hook/override_log.py:36this_chunks: reflog semantics and a three-level conditional at :96skills/larv-review/SKILL.md:64"Round cap": rules pointing at each other, exceptions on exceptionsbin/loop_claude/cycle/judge.py:127run_judge: two quote checks with different tolerancebin/shell_command.py:31segments: heredoc and comment regexes and a redirect flag at oncebin/larv_review/plan.py:8scoring_run_due: plan before and after, shipped IDs and tag state
Mid A, hardest first:
skills/larv-review/SKILL.md:64"Round cap": a dozen interlocking conditions simulated in the headbin/larv_review/record.py:14record_ruling: branch order carries meaning, terms defined only in the skillsbin/loop_claude/cycle/judge.py:16JUDGE_RULES: the prompt assembled from three filesbin/comprehension_history.py:110seat_places: a parser with a flag reset by any unindented linebin/rule_hook/override_log.py:86print_overrides: reflog semantics and a chained conditionalbin/shell_command.py:31segments: regex-heavy, no inline samples
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:29"Passes": a rule engine in prose, its mechanised half unmarkedbin/larv_review/record.py:14record_ruling: four exits over two parallel vote mapsbin/loop_claude/cycle/judge.py:16JUDGE_RULES: the judge's contract built across three filesbin/rule_hook/override_log.py:86print_overrides: reflog membership and a three-level conditionalbin/shell_command.py:31segments: five concerns in one pass, two tools depend on itbin/transcript_archive/write.py:13archive: one invariant held across three files
Inherited architect, would restructure first:
bin/larv_review/rounds.py:8round_counts: the round cap is enforced only by prose across five placesbin/larv_review/git.py:21default_branch: one of two default-branch rules and five git runnersbin/loop_claude/loop/main.py:8module docstring: the rule-folder list hand-written in four placesbin/rule_hook/rules.py:73Verdict: one name over three concepts, likeEntryandrulingbin/rule_hook/rules.py:108git_subcommand: a second commit detector besidereading_cost/measure.py:9COMMIT
2026-10-13T06:49:43Z, PR #245 at a796bb3, against 4cc9953
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | same | same | same |
| Inherited architect | same | worse | better | same | same |
Mid A, decided by:
- Locality:
bin/git_command.py:16commits: one shared parser replaces reading-cost's lexer and regex, at the cost of a parser shared with rule-hook - Locality, Self-sufficiency:
skills/larv-review/standards/prose.md:3"Prose": a full copy beside CLAUDE.md's core, against small single-topic files - Navigation:
CLAUDE.md:38"Topic rules": three homes for a rule, all listed in the README map - Shape:
bin/rule_hook/git.py:11git: a bounded package, butrule_hook/git.pybesidegit_command.pyand some vague names
Inherited architect, decided by:
- Locality:
CLAUDE.md:21"Prose": two spellings of the prose rules, already worded differently - Locality:
agents/prose-reviewer.md:9"The bar": the bar is assembled from three sources and path globs - Locality:
bin/loop_claude/project/records.py:16read_records: any file added to the rule folders lengthens the judge's records - Locality:
bin/rule_hook/override_log.py:36this_chunks: reflog, default branch and logged branch held at once - Shape:
bin/reading_cost/measure.py:43commit_time: reading-cost and rule-hook share one parser, one concept with one implementation - Shape:
bin/rank_todo.py:114check_failures: --check pulled out so rule-hook reuses it
2026-10-13T09:29:12Z, PR #254 at 2284780, against 4cc9953
Ruling: worse
Fell: Locality. Under the floor of 7.0: Locality 6.50, so the same ruling counts as worse.
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | same | same |
| Inherited architect | same | worse | same | same | same |
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6.5 | 6.5 | 6.5 | 6.5 |
| Mid A | 7.5 | 7 | 7 | 6.5 | 7 |
| Maintainability senior | 7.5 | 6.5 | 6.5 | 7 | 7 |
| Inherited architect | 7 | 6 | 6 | 6.5 | 6.5 |
| Mean | 7.25 | 6.50 | 6.50 | 6.63 | 6.75 |
Mid A, decided by:
- Locality:
bin/todo_table.py:62parse_row: a C-less row gets a blank cell spliced in by position, and whether a table without C is valid is decided in three files that disagree on purpose - Locality:
bin/loop_claude/project/records.py:9HUMAN_RULE_DIRS: the judge's rule-file text is now built from tuples across three modules and an unnamed repo check - Locality:
bin/rule_hook/override_log.py:118default_branch: reuses the name oflarv_review/git.py'sdefault_branchwith a different contract
Inherited architect, decided by:
- Locality:
bin/score_outcomes/outcomes.py:6round_findings: the round-commit title is a contract held in SKILL.md prose,ROUNDandFINDINGS; a slightly different title silently reads as findings unknown - Locality:
bin/todo_table.py:62parse_row: the rule for the pre-C table is spread over todo_table, rank_todo and larv_review/plan, which disagree on purpose - Locality:
bin/rule_hook/override_log.py:36this_chunks: which overrides belong to the chunk depends on reflog and ancestry state the code does not hold - Locality, Shape:
bin/score_outcomes/main.py:41imports: one tool reaches into another tool's package for its git helpers
Junior A, hardest first:
bin/score_outcomes/outcomes.py:42guesses: a second todo-table parser besidetodo_table.itemsbin/larv_review/git.py:8git: one of four look-alike git wrappers with different contracts
Mid A, hardest first:
bin/score_outcomes/outcomes.py:9FINDINGS: round titles parsed inrounds.py, here and prose in SKILL.mdbin/score_outcomes/outcomes.py:21Outcome: one of three unrelatedOutcomes
Maintainability senior, hardest first:
bin/score_outcomes/outcomes.py:42guesses: a third items-table parser with its own cell splitterbin/score_outcomes/main.py:93measure: decisions threaded between git calls
Inherited architect, would restructure first:
bin/score_outcomes/outcomes.py:9FINDINGS: the round-title grammar lives in SKILL.md,ROUNDand herebin/larv_review/git.py:21default_branch: one of twodefault_branchs and five git runners
2026-10-13T10:03:42Z, PR #254 at 5b6c497, against 4cc9953
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | same | same |
| Inherited architect | same | worse | same | same | same |
Mid A, decided by:
- Shape:
bin/git_repo.py:8git: four modules carry "git" in their path, and two define agitwith different signatures - Shape:
bin/todo_table.py:30items: one name now covers three outcomes and two header shapes, each caller reading None differently
Inherited architect, decided by:
- Locality:
bin/todo_table.py:30items: accepts a table without C thatrank_todo.pyread_scoringrejects, andrender_itemrelies on that rejection having run first - Locality:
bin/larv_review/plan.py:46plan_reasons:items(before) or []turns None into an empty list and relies on a Failure being truthy - Locality:
bin/loop_claude/project/records.py:10HUMAN_RULE_DIRS: the judge reads every file in two folders, shrinking the transcript it sees - Locality:
bin/loop_claude/cycle/judge.py:41JUDGE_RULES: the judge's rule text is built across three files - Locality:
bin/rule_hook/override_log.py:36this_chunks: which overrides list depends on reflog, branch name and default-branch state
2026-10-13T10:23:39Z, PR #255 at 459df0a, against 4cc9953
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | same | better | same | worse | better |
Mid A, decided by:
- Locality, Overall:
skills/larv-review/SKILL.md:99"Comprehension check": the check is one self-contained subsection that runs once in §5 and blocks nothing; at4cc9953Ratchet, Floor, Fell, Fresh, Rescore and the round cap referred to each other - Locality:
bin/larv_review/record.py:14record_ruling: one straight path now; at4cc9953it branched four ways on a fall and a scores file, withread_fall_scores,refuse_recordedandfall_rulingto follow - Locality:
bin/reading_cost/measure.py:43commit_time: finds the worker's first commit through the sharedgit_command.commitsparser rule-hook uses, in place of its own regex and tokenizer
Inherited architect, decided by:
- Locality, Overall:
skills/larv-review/SKILL.md:99"Comprehension check": runs once, blocks nothing, under its own heading; at4cc9953it forced cross-references into Round cap, Later rounds and No rerun - Locality:
bin/larv_review/record.py:14record_ruling: reads one votes file and appends one ruling; at4cc9953a two-call state machine withFell:and a second file - Locality:
bin/reading_cost/measure.py:43commit_time: one parse of a command shared with rule-hook, in place of a regex and a separate lexer - Self-sufficiency:
bin/larv_review/ruling_base.py:7measured_commit: the docstring still calls the base the ratchet's base and speaks of falls nothing in the code defines - Self-sufficiency:
bin/comprehension_history.py:43Entry: docstrings and theplaceshelp describe the scores a fall called into a ruling, which needs the project's history open to parse - Self-sufficiency:
AGENTS.md:5decision titles: one says a project under the floor is fixed at the lowest dimension, others that the check blocks nothing, so a reader cannot tell whetherFloor:means stop - Self-sufficiency:
skills/larv-review/standards/quality.md:6quality bar: points at §3 "Architecture" for the comprehension check, which now has its own subsection - Locality:
bin/rule_hook/override_log.py:118local_default_branch: a second default-branch resolver besidebin/git_repo.py:21default_branch, neither naming the other
2026-10-13T14:48:28Z, PR #255 at a12b685, against 4cc9953
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | same | better | same | same | better |
Mid A, decided by:
- Locality, Overall:
skills/larv-review/SKILL.md:101"Comprehension check": runs once in §5, reports no findings and blocks nothing; at4cc9953"Fresh" tied it into the passes, and "Round cap", "Re-entry" and "Suspended" pointed back to it - Locality, Overall:
bin/larv_review/record.py:14record_ruling: parse, count, split or rule, append; at4cc9953it branched on the scores file and a fall, throughread_fall_scoresandfall_rulingin three modules - Locality:
bin/larv_review/ruling_base.py:4measured_commit: the newest change adding any entry reads on its own; at4cc9953basedepended on a better ruling, heading times and afirst_rulingfallback - Locality:
bin/reading_cost/shell_reads.py:21shell_reads: shell parsing moved tobin/shell_command.py:20simple_commands, shared with rule-hook
Inherited architect, decided by:
- Locality, Overall:
bin/larv_review/record.py:14record_ruling: runs in a straight line with no mode flag onread_seats; at4cc9953four branches on a fall and a scores file - Locality, Overall:
bin/larv_review/ruling_base.py:4measured_commit: one loop, andshow_ruling_baseno longer reads the history; at4cc9953a two-step lookup throughfirst_ruling - Locality, Overall:
skills/larv-review/SKILL.md:101"Comprehension check": runs once, blocks nothing, no pass; at4cc9953"Fresh" made a prose fix in §3 change what §5 must do - Locality:
bin/comprehension_history.py:126worse_votes: recovers votes by re-parsing table rows by cell count, relying on the votes table coming first, which matters only for legacy rulings - Locality:
bin/rule_hook/git.py:11git: same name asbin/git_repo.py:8gitwith a different contract, beside two more git-named units
2026-10-13T15:24:20Z, PR #257 at d471377
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6 | 6 | 6 |
| Mid A | 7 | 7 | 6.5 | 6.5 | 6.5 |
| Maintainability senior | 7 | 6.5 | 6.5 | 7.5 | 7 |
| Inherited architect | 7 | 7 | 6 | 6.5 | 6.5 |
| Mean | 7.00 | 6.63 | 6.25 | 6.63 | 6.50 |
Junior A, hardest first:
skills/larv-review/SKILL.md:63"Round cap" and:74"Re-entry": seven interlocking rules on reruns, caps and exceptions, each pointing elsewhere; what a fix to a stability 🟡 that touches a comment reruns stayed unclearbin/comprehension_history.py:106latest_places: a 70-word docstring, an index rule resting on a pre-PR #255 history quirk, andworse_voteskeeping the first row bysetdefaultbin/loop_claude/cycle/judge.py:16JUDGE_RULESand:127run_judge: behaviour in a prose prompt, and three text normalisations held at oncebin/rule_hook/override_log.py:36this_chunks: reflog and branch-name rules that need git knowledgebin/shell_command.py:31segments: dense regexes and a second lexing path for unbalanced quotesbin/loop_claude/cycle/states.py:69transition: the quoted-answers counter threads through three statesbin/larv_review/plan.py:9scoring_run_due: "landed", "next release" and "emptied" mix two plansbin/transcript_archive/write.py:13archive: three invariants on folder naming and write order
Mid A, hardest first:
skills/larv-review/SKILL.md:63"Round cap": about 30 interlocking rules across the trigger table, the fix clauses, the cap and later rounds; only the count is enforced in codebin/comprehension_history.py:106latest_places: parses its own markdown by label regexes, resting on table order and a legacy casebin/loop_claude/cycle/judge.py:16JUDGE_RULES: the prompt is assembled from constants in three modulesbin/rule_hook/override_log.py:36this_chunks: reflog semantics trusted, not verifiable by readingbin/shell_command.py:31segments: a hand-rolled tokenizer two unrelated tools depend onbin/larv_review/ruling.py:8votes: the vote-counting rule spread over four modules and an 80-line help, with a ruling decision inside a parserbin/rank_todo.py:378rank_details: regex split and string splicing whose newline round-trip needs running to trustbin/transcript_archive/write.py:13archive: the why lives inloop_claude/loop/main.py's docstring
Maintainability senior, hardest first:
skills/larv-review/SKILL.md:63"Round cap" and:74"Re-entry": a state machine written as prose, re-derived on each readbin/comprehension_history.py:106latest_places: dense comprehensions, a seat name recovered by splitting a label, and table-order dependencebin/loop_claude/cycle/judge.py:16JUDGE_RULESand:127run_judge: one prompt spliced from three files, and two misquotes treated differentlybin/rule_hook/override_log.py:36this_chunks: four branches over reflog behaviourbin/reading_cost/measure.py:27measure: timestamps compared as strings with the format assumption unstatedbin/rule_hook/rules.py:49OVERRIDE: a lazy reason with a lookahead and a suffix strip, explained nowhere in the codebin/larv_review/plan.py:9scoring_run_due: the "work has landed" rule, written three timesbin/loop_claude/loop/reload.py:50exec_if_edited: editing anybin/**/*.pyre-execs a running loop
Inherited architect, would restructure first:
bin/round_titles.py:5PASSES: repeats the pass table atskills/larv-review/SKILL.md:33with no check tying them, and omits the auto-reviewerskills/larv-review/SKILL.md:71"Cap": the round cap is prose; no code stops a fourth round, so the 3am path takes 20 to 40 minutesbin/git_repo.py:8git: one of five git runners with four signaturesbin/larv_review/git.py:6history_changes: a module named git that holds history logicbin/reading_cost/measure.py:16Reading: "places" also names comprehension places, and "record" and "outcome" each carry three meaningsbin/loop_claude/cycle/judge.py:16JUDGE_RULES: says "the agent" where the code says "worker", and restates the working loop uncheckedbin/loop_claude/loop/reload.py:38fingerprint: covers everybin/**/*.py, so editingstatusline.pyreloads the loop
2026-10-14T21:34:44Z, PR #262 at f52228d, against e17168e
Ruling: worse
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | worse | worse | worse | same | worse |
| Inherited architect | worse | worse | worse | same | worse |
Mid A, decided by:
- Locality, Overall:
bin/loop_claude/routing.py:10ROUTED: a chunk's rung survives a rerun only as prose thatRung.__str__andRoute.__str__write and a regex in another package parses back; rewording one string, or a new outcome kind, silently breaks resume three files away - Locality, Shape:
bin/loop_claude/loop/main.py:164routed:run_loopnow threads aRoutethroughLoopRunwith two lazy lambdas, and following one step-up means holdingstepped, the judge'sbundled,after_verdictandarchive_steptogether - Locality:
bin/loop_claude/loop/reload.py:41decoded_as: the flat key and type check became a recursive decoder, and the comment aboveLoopRunstill says a field change stops a running loop - Navigation:
bin/model_routing/ladder.py:1LADDER: routing is split across two modules named for routing plus five helpers inmain.py, and README → Tools still saysbin/'s shared modules serve more than one tool - Shape:
bin/loop_claude/cycle/judge.py:54Reply.bundled: the name says neither R + C nor highest, andkeptandroutedmake a reader open their bodies
Inherited architect, decided by:
- Locality, Navigation:
bin/loop_claude/routing.py:10ROUTED: a string contract spread over four files,Rung.__str__,Route.__str__,NAMES_A_WORKERand the regex; a reword makesroute_fromreturn None and the chunk routes afresh with no error - Shape, Navigation:
bin/loop_claude/loop/main.py:129run_loop: once a flat sequence, it now threads an anchor, a route, a before and after pair and two lambdas, and its imports grew from 13 modules to 21 - Shape:
bin/loop_claude/loop/main.py:180archive_step:archiveis a function and a Path, andproject_archivea function and a parameter, a few lines apart - Locality, Shape:
bin/loop_claude/loop/reload.py:41decoded_as: a recursive decoder with a sentinel, and the comment aboveLoopRunhalf-contradicts the one inside it - Shape:
bin/loop_claude/routing.py:59stepped: does two jobs, raising to a bundled item's rung and counting towards a one-rung step, and its name covers only the second
2026-10-14T22:22:01Z, PR #264 at 17e1ebe, against 8f7b0c5
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | same | same |
| Inherited architect | same | worse | same | same | same |
Mid A, decided by:
- Locality:
bin/loop_claude/cycle/prompts.py:27off_the_projects_rung: the prompt text now depends on how the worker's rung compares with the project's throughhighest, and why a chunk with no item needs no note rests onfirst_routein another module - Locality:
bin/loop_claude/routing.py:54keptand:59stepped: the two now use different floors, so confirming a route below the project's rung survives a reload means checkingroutedinloop/main.pyagainst both - Locality:
bin/loop_claude/cycle/run.py:36Cycle.anchor: a second rung travels besiderouteinto every prompt call, each with its None case
Inherited architect, decided by:
- Locality:
CLAUDE.md:55"Working with the codebase": which model a subagent gets now spans this prose rule,prompts.py'soff_the_projects_rungandladder.py'sRung.agent_args; that reviewers above the project's rung inherit the worker's rung is said in neither file - Locality:
bin/model_routing/ladder.py:16Rung.agent_args: the family name comes fromself.model.split('-')[1], tying the ladder's ids to the Agent tool's vocabulary with nothing checking it - Locality:
bin/loop_claude/routing.py:54keptand:59stepped: whether a rerun resumes now depends on the R + C the archived route line holds, and the order of the checks is more to hold
2026-10-14T23:10:29Z, PR #265 at ac41116, against 8bc0d88
Ruling: worse
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | worse | same | worse |
| Inherited architect | worse | worse | worse | worse | worse |
Mid A, decided by:
- Locality:
bin/loop_claude/loop/main.py:186run_loop: every cycle now ends with a cross-project check that counts every project's cycles, takes a lock and may hand the terminal to a claude in ~/.claude, so the loop no longer reads alone - Locality:
bin/loop_claude/loop/routing_check.py:26new_callsandbin/routing_figures/archive.py:59last_measured: whether a measurement is due is read back by regex from the markdownrenderwrites, so rewording it silently breaksdue; the raised-calls file's lifecycle ends in a manual rule in AGENTS.md - Locality:
bin/routing_figures/measure.py:130spend: repeatsreading_cost/measure.pytokens_in's dedupe-by-message rule, adding output tokens, so a counting change has two places to land - Shape:
bin/routing_figures/measure.py:74measured: dense single expressions buildingReportsand a six-argumentChunkFigures, and nested comprehensions inrender - Shape:
bin/routing_figures/measure.py:43Reports:searchesholds characters of tool results, not a count of searches; generic namescalls,due,add,totals,kept
Inherited architect, decided by:
- Locality, Self-sufficiency:
bin/loop_claude/loop/routing_check.py:46raise_calls: how a routing call is raised, deduplicated and retired is split across the code, the prompt and a working rule in AGENTS.md about cited stamps and a hand-appended Reset - Locality:
bin/routing_figures/measure.py:193callsandbin/routing_figures/archive.py:124raised: a call's identity is its English sentence, and the measured cycle is read back from rendered markdown, so the report's prose is load-bearing state - Locality:
bin/loop_claude/loop/main.py:186run_loop: each project's loop now runs a cross-project step holding a lock and scanning the whole archive, so one loop depends on what other loops wrote - Shape:
bin/routing_figures/measure.py:74measured: one 234-line module holds six concerns, andmeasured,Measuredandmeasure_archivesit side by side;keptandcallseach name two things - Navigation:
bin/model_routing/route.py:11Route: moved out of loop_claude while still carrying loop-only state, so "routing" now spans five places - Self-sufficiency:
bin/routing_figures/measure.py:130spend: the docstring says it counts astokens_indoes, but it adds output tokens
2026-10-15T07:36:58Z, PR #266 at 6cc3ded, against 8412a48
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | better | same | same | same | better |
Mid A, decided by:
- Locality:
bin/transcript_archive/location.py:9FILE_SUFFIX, withoutcomes_ofandtranscript_of: the one place that maps a transcript copy to its outcomes file and back, where the base wrote the suffix and the inverse step out in write.py, read.py and routing_figures/archive.py - Locality:
bin/transcript_archive/read.py:31archived_outcomesandchunk_outcomes: replace the raw globs routing_figures made, so it no longer knows the archive's folder depth - Locality:
bin/transcript_archive/outcomes.py:6TASK,INTERRUPTED,ANSWERED,TASK_AT,ANY_VERDICT_AT: writer and readers share one set of constants, so the outcomes format reads and changes in one file - Locality:
bin/loop_claude/routing.py:36item_rung:keptandjudge_rungshare one named helper where the base held the rule in an inline conditional - Locality:
bin/loop_claude/cycle/judge.py:127judge_onceandjudge_args: one judge run and its flags stand alone, so the replay reuses them without copying
Inherited architect, decided by:
- Navigation, Overall:
bin/transcript_archive/outcomes.py:22JUDGE,interrupted,answered: each archived verdict carries aJudge:line, so a reader paged about a wrong ruling reads which rung judged it from the outcomes file - Navigation, Locality, Overall:
bin/transcript_archive/location.py:9FILE_SUFFIX,outcomes_of,transcript_of, withbin/transcript_archive/read.py:31archived_outcomes: the archive layout lives in one module, thougharchived_outcomesstill derives the key frompath.parent.parent - Locality:
bin/judge_replay/main.py:124first_reply: re-implements the loop's retry count besideretried, two places that can drift - Locality:
bin/project_key.py:9checkout_of: a guessing inverse ofproject_keythat maps a project archived from/xto~/code/xwhen that folder exists, silently - Shape:
bin/judge_replay/main.py:31imports: one launcher reaching into fiveloop_claudeinternals, balanced byjudge,judge_onceandjudge_argsinbin/loop_claude/cycle/judge.py:123now three units with truthful names - Locality:
bin/loop_claude/routing.py:51judge_rung: derives its own anchor from the judge model whileCycle.anchoris the worker's, so "anchor" means two rungs in one cycle when the models differ - Self-sufficiency:
bin/loop_claude/routing.py:54judge_rungdocstring: points into the plan at item 136, andJUDGE_REPLAYEDandREFERENCE_EFFORTrest on a past replay and a past settings value the code cannot show
2026-10-15T14:08:19Z, PR #273 at f13447f, against b281392
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | better | same | same |
| Inherited architect | same | worse | same | same | same |
Mid A, decided by:
- Shape:
bin/model_routing/ladder.py:52floored: names the rule "the rung R + C picks, never below the project's rung" besidepicked, for two callers, where the base built it inline injudge_rungfromhighestanditem_rung
Inherited architect, decided by:
- Locality:
rules/todo-items.md:9"Plan only on the project's rung or above": the planning rule spans the rule file,CLAUDE.md:55,bin/subagent_rung/main.py's help andladder.floored, with no one file holding it whole - Locality:
bin/loop_claude/cycle/prompts.py:34off_the_projects_rung: the prompt spells out another tool's command line,subagent-rung --plan, with nothing tying the two, and its docstring points torouting.first_routein another module - Locality:
bin/loop_claude/routing.py:61judge_rung: reads in one line and drops the two-way reference withitem_rung, but writes the no-item branch a second time - Shape, Self-sufficiency:
bin/model_routing/ladder.py:52floored: a truthfully named helper, offset bybin/subagent_rung/main.py:23agent_args's booleanplanningand position-sensitive--planparsing
2026-10-15T14:39:39Z, PR #274 at 1431cc6, against 2190ec8
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | same | worse | same |
| Inherited architect | same | worse | worse | worse | worse |
| Maintainability senior | same | worse | same | worse | worse |
| Junior A | same | same | same | worse | same |
Mid A, decided by:
- Self-sufficiency:
bin/routing_figures/measure.py:106escalations: the figures count escalations by a bare trigger number, and only the prose list inCLAUDE.mdsays what each number is - Self-sufficiency:
bin/loop_claude/routing.py:43escalation_rungs: the docstring explains the cap by naming who set it, a reason only to someone who knows the decision
Inherited architect, decided by:
- Locality:
bin/loop_claude/cycle/prompts.py:44escalation: the marker's<n>:format is typed in the prompt and again inmeasure.py's regex, and a mismatch makes the figures print None with no error - Locality:
bin/loop_claude/routing.py:43escalation_rungs: the upper bound counts fromitem_rungand the lower fromroute.rung, so a step-up shrinks or empties the window, explained only by naming who set it - Locality, Shape:
bin/loop_claude/cycle/prompts.py:46escalation: the past-the-cap branch is a third case the docstring omits, and prompts.py now imports policy fromloop_claude.routing - Shape:
bin/model_routing/route.py:8ESCALATION_MARKER: the module is about a chunk's Route as archived, but the marker belongs to the subagent-prompt contract - Shape, Self-sufficiency:
CLAUDE.md:60"Working with the codebase": "escalation" already named the reviewer-to-reviewer 🔀 handoff inskills/larv-review/SKILL.md:80, and now also names handing a task one rung up - Self-sufficiency:
bin/routing_figures/measure.py:258triggers:routing-figures.mdprintstrigger 2: 3, and no code names what each number means
Maintainability senior, decided by:
- Locality:
bin/loop_claude/cycle/prompts.py:40escalation: when the sentence appears is split between its own top-rung guard andescalation_rungs's empty slice in another module - Locality:
bin/loop_claude/cycle/prompts.py:19fresh: one plain conditional became a nested one with anlstrip, androuted_byalso appends the sentence, so both paths need tracing to rule out a double - Self-sufficiency, Locality:
bin/routing_figures/measure.py:106escalations: counts bare trigger numbers that onlyCLAUDE.md's numbered list defines - Self-sufficiency:
bin/loop_claude/routing.py:43escalation_rungs: the cap's base is explained by who set it, and the slice takes decoding
Junior A, decided by:
- Self-sufficiency:
bin/routing_figures/measure.py:23ESCALATION: trigger numbers mean something only inCLAUDE.md:60's list, and renumbering it would silently change old figures - Self-sufficiency, Shape:
CLAUDE.md:60"Working with the codebase": "trigger" now names an escalation condition as well as the review-pass condition Vocabulary defines - Self-sufficiency, Locality:
bin/loop_claude/cycle/prompts.py:40escalation: the past-the-cap branch is missing from the docstring, and the off-ladder guard differs fromescalation_rungs's
2026-10-15T19:39:18Z, PR #276 at f643a14, against eb23048
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | worse | same | worse |
| Inherited architect | same | same | worse | better | same |
| Maintainability senior | same | same | worse | same | same |
| Junior A | same | same | worse | same | same |
Mid A, decided by:
- Shape, Locality:
bin/rung_replay/search.py:44replayed: generic overCase,ItemandReplywith areproducedcallback returning a list of items, worked throughheld,aimandgot; for the judge the list is only[kind]or[], so the reader decodes an abstraction built for two consumers before following one - Locality:
bin/judge_replay/replay.py:70report: rebuilds the unreproduced verdicts by callingreproducedagain onreplay.on_reference, so it is right only whilemain.pypasses that same function toreplayed - Shape:
bin/stability_replay/main.py:83main:foundnames aRound, aset[int]and a list in one file, and one call packs the reference, the candidates, a closure overcodesand an inline lambda - Shape:
bin/rung_replay/command.py:14project_rung_or_exit: returnsRung | int, the int an exit code, so each caller testsisinstance(anchor, int)
Inherited architect, decided by:
- Shape:
bin/rung_replay/search.py:44replayed: generic over its type variables, so a reader works out whatreproducedreturns before the search reads; it was a concrete comparison of kinds - Shape:
bin/judge_replay/replay.py:65reproduced: returns a list of zero or one kinds where a yes or no would do, to fit the generic shape - Shape:
bin/stability_replay/main.py:83main:foundis aRoundin one lambda and aset[int]in the next, and an accumulator list elsewhere in the file - Self-sufficiency:
bin/rung_replay/search.py:52replayed: the docstring says why the search goes down from the highest rung and stops at the first miss, which the cheapest-first search assumed unsaid - Self-sufficiency:
bin/stability_replay/main.py:120code_of: says what it guesses and what a wrong guess costs, so no record is needed beside it
Maintainability senior, decided by:
- Shape:
bin/rung_replay/search.py:44replayed: generic asReplay[Case, Item, Reply]with areproducedcallback returning a list, which judge-replay fits only with a list of one or none and stability-replay with an inline adapter; following onereportline takes three files and the generics - Shape:
bin/stability_replay/main.py:66main: one file walks the archive, rebuilds the reviewed commits from two git logs, clones, reviews and matches, with long lines packing thereplayedcall, a nested conditional overfirst_replyand a starred conditional checkout command
Junior A, decided by:
- Shape:
bin/rung_replay/search.py:37Replay: the search is generic over three type parameters with two callbacks and a label, so a reader works out whatItemandReplyare for each caller before the loop makes sense - Shape:
bin/judge_replay/replay.py:65reproduced: returns[kind]or[]only to fit the generic shape, andreportrebuilds the unreproduced list by calling it again - Shape:
bin/rung_replay/command.py:1command: the name says little, and the module mixes argument parsing, a models-file lookup, rung arithmetic and stderr exits
2026-10-15T20:33:16Z, PR #278 at 0f71e33, against ddb86cb
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | better | better |
| Inherited architect | same | same | worse | better | same |
| Maintainability senior | same | same | worse | better | same |
| Junior A | same | same | same | better | better |
Mid A, decided by:
- Locality:
bin/loop_claude/routing.py:38kept_rung: one rule now raises a route's rung to item_rung for bothkeptandstepped, where the base handled the off-ladder case in two places with two different targets - Locality, Self-sufficiency:
bin/loop_claude/routing.py:55escalation_cap: the escalation step and the prompt's offered rungs share one named cap, whose docstring gives the reason it counts from item_rung instead of naming who set it - Self-sufficiency:
bin/loop_claude/cycle/judge.py:149run_judge: a docstring lists the three quotes, what each is checked against and what a misquote does, where the base had one comment covering only the quoted answer - Self-sufficiency:
bin/loop_claude/routing.py:63stepped: the docstring says why it takes the verdict's fields and not the verdict, an import cycle through prompts
Inherited architect, decided by:
- Shape:
bin/loop_claude/routing.py:63stepped: the starting rung now takes three helper hops,taken_in,kept_rung,highest, where the base had two inline lines, and the off-ladder handling is said only inkept_rung's docstring - Shape:
bin/loop_claude/routing.py:78taken_in: used once, its name does not say what it returns, andkept(route, anchor)andkept_rung(anchor, route)take their arguments in different orders - Self-sufficiency:
bin/loop_claude/routing.py:55escalation_cap: the cap's reason is now in the code, and it is computed once forescalation_rungsandstepped - Self-sufficiency:
bin/loop_claude/cycle/judge.py:149run_judge: the docstring lists all three quote checks and why no misquote but the evidence's is a failure, which answers the 3am question without opening anything else - Self-sufficiency:
bin/loop_claude/routing.py:66stepped: the docstring explains the import cycle that keeps the verdict type out of routing
Maintainability senior, decided by:
- Self-sufficiency:
bin/loop_claude/cycle/judge.py:149run_judge: the docstring lists the three quote checks, what each is checked against and what a misquote does, where the base had one comment covering only the answer - Self-sufficiency:
bin/loop_claude/routing.py:63stepped: the docstring states the import cycle that keeps the verdict type out, andescalation_capsays why the cap counts from item_rung - Shape:
bin/loop_claude/routing.py:11ESCALATION_CAP: differs only in case fromescalation_capat:55, an int against a function returning aRung, and its comment ends in a stray "yes,." - Shape:
bin/loop_claude/routing.py:78taken_in:steppedno longer reads top to bottom; it chasestaken_in,kept_rung,item_rungandpicked, and the off-ladder rule lives only inkept_rung's docstring
Junior A, decided by:
- Self-sufficiency, Overall:
bin/loop_claude/cycle/judge.py:149run_judge: the docstring lists the three quote checks, how strictly each is checked and what a misquote does, where the base made the reader work out why grounded and verbatim differ - Self-sufficiency:
bin/loop_claude/routing.py:55escalation_cap: the limit is its own named function whose docstring says why it counts from item_rung, andsteppednamescounted_outandescalates
2026-10-15T21:39:57Z, PR #280 at 9af5ee7, against 545f3c3
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | same | same | worse | same | same |
| Maintainability senior | same | better | same | same | same |
| Junior A | same | better | same | same | better |
Mid A, decided by:
- Locality, Overall:
bin/todo_table.py:64items_table: reads the items table once and takes each cell by column name, where the base needed a second header constant thatitems,read_scoringandread_itemseach had to know - Locality, Overall:
bin/todo_table.py:81waiting: names the rule for skipping an item once, used byrouted_item, where the base wrote it inline - Locality:
bin/loop_claude/routing.py:84judge_rung: callsitem_rung, so the judge's and the worker's rung share one definition
Inherited architect, decided by:
- Shape:
bin/rank_todo.py:311place: five parameters, recursion, a set rebuilt per call, and a name that hides that it also finds loops and prunesblocked_by - Locality:
bin/rank_todo.py:194rank: the rule that a blocker that left the table no longer counts is worked out here, inplaceand intodo_table.waiting - Locality:
bin/rank_todo.py:274read_items: the Blocked by checks are spread overparse_row,read_items,blockers_firstandplace - Locality, Shape:
bin/todo_table.py:64items_table: cells read by column name, with the C check beside the column, where the base split the header text by position - Locality:
bin/loop_claude/routing.py:16routed_item: asks the purewaiting, where the base leaned on an order rank_todo kept - Shape:
bin/rank_todo.py:187rank: the two-line header replacement keepsfound.startright only because header and divider are two lines, and the module's fan-out grows past 25 functions
Maintainability senior, decided by:
- Locality:
bin/todo_table.py:107parse_row: reads each cell by the column name in the table's own header, where the base found the table by exact header strings kept in mind across two modules - Locality:
bin/rank_todo.py:247read_items: the C and unknown-column checks sit where the table is read, where the base checked raw lines in the formula reader - Locality:
bin/rank_todo.py:311place: a partial offset; dropping a blocker that left the table happens inplace,blockers_firstandrank, andplacethreads five parameters - Locality:
bin/loop_claude/routing.py:16routed_item: defers to the named, order-freetodo_table.waiting, where the base leaned on rank_todo's order
Junior A, decided by:
- Locality, Overall:
bin/todo_table.py:64items_table: one function finds the table, checks repeated and missing columns, and rows are read by column name, where the base kept two header variants in step - Locality:
bin/rank_todo.py:247read_items: the table's shape checks sit where the table is read, where the base rejected a table in the scoring reader - Locality:
bin/loop_claude/routing.py:16routed_item: the waiting rule is one named function,todo_table.waiting - Locality:
bin/loop_claude/routing.py:84judge_rung: builds the judge's rung fromitem_rung, one definition fewer to keep in step - Locality:
bin/rank_todo.py:311place: the counterweight; a two-stage ranking with a five-parameter recursion, and the dropped-blocker rule written three times
2026-10-15T21:39:57Z, PR #280 at 9af5ee7
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6 | 6 | 6 |
| Mid A | 7 | 6.5 | 6.5 | 6 | 6.5 |
| Maintainability senior | 7 | 6 | 6 | 7 | 6.5 |
| Inherited architect | 7 | 7 | 6 | 7 | 6.5 |
| Mean | 7.00 | 6.38 | 6.13 | 6.50 | 6.38 |
Junior A, hardest first:
bin/loop_claude/routing.py:64stepped: six one-line helpers held at once to learn the next cycle's rung, each docstring defined by the othersbin/loop_claude/cycle/prompts.py:40escalation: the worker's prompt glued from space-led fragments over several cases, with no example of the final stringbin/comprehension_history.py:106latest_places: hand-shaped markdown read back by regexes, an in-a-list flag and history from before PR #255bin/loop_claude/cycle/judge.py:149run_judge: three quote checks with three consequences, and a JUDGE_RULES prose wall built from constants in other modulesbin/transcript_archive/outcomes.py:53workers: a chunk resumes its rung by regex-parsing archive prose, with the verdict strings repeated by handbin/routing_figures/measure.py:77measured: timestamps cut to 19 characters, tools carried between iterations, totals into four countersbin/stability_replay/rounds.py:70reviewed_path: guesses a removed worktree from absolute paths with a first-prefix-gone rulebin/rank_todo.py:311place: recursive placement with a loop-detecting path and a placed set rebuilt per call
Mid A, hardest first:
bin/loop_claude/routing.py:64stepped: five functions pick a rung from four reference points, and the why is a pointer to the decision logbin/loop_claude/cycle/prompts.py:40escalation: conditional f-string fragments relying on leading spaces and lstrip, output unknowable from the codebin/transcript_archive/outcomes.py:53workers: resume rests on a parsed Worker line, and NAMES_A_WORKER repeats the verdicts as literalsbin/comprehension_history.py:106latest_places: a five-clause docstring, table-order dependence and an implicit line state machineskills/larv-review/SKILL.md:63"Round cap": the fourth-round rule lives in cross-referenced prose that no code enforcesbin/routing_figures/measure.py:77measured: dense one-liners and a timestamp format assumed silently inrung_atbin/rule_hook/override_log.py:36this_chunks: which overrides belong to the chunk rests on reflog semanticsbin/rank_todo.py:311place: recursive blocker placement with a loop-detecting path, resolved after one reread
Maintainability senior, hardest first:
bin/loop_claude/routing.py:64stepped: seven functions over three inputs, and the never-down and two-above rules emerge only from their compositionbin/loop_claude/cycle/prompts.py:29off_the_projects_rung: prompt fragments and rung arithmetic over three branches in each of three functionsbin/loop_claude/loop/main.py:1"loop-claude --help": a 120-line contract, the only end-to-end routing story, with no pointers to the modules it maps ontoskills/larv-review/SKILL.md:63"Round cap": which passes rerun after a fix spans the pass table, the fix clauses and the capbin/model_routing/ladder.py:19Rung.__str__: the archive's rung format is written and parsed across four files, and a reworded string breaks resumebin/routing_figures/measure.py:77measured: tokens charged to the previous message's tools, and timestamps cut to compare with stampsbin/shell_command.py:31segments: a hand-rolled splitter whose words after a redirect target open a new segmentbin/comprehension_history.py:106latest_places: a four-clause docstring, a seat name split from a label and a votes table assumed first
Inherited architect, would restructure first:
bin/round_titles.py:4CAP: the round cap lives here for its regex's other readers, and no program enforces it;SKILL.md"Round cap" does in prosebin/loop_claude/effects.py:18Effects: the process shell,Stopand the models-file reader that three other tools reach into loop_claude forbin/git_repo.py:8git: one of seven ways to run git, alongsidemetric_bench/main.py:150,stability_replay/main.py:202andreading_cost/main.py:114bin/score_outcomes/outcomes.py:156outcomes_of: shares its name and module name withtranscript_archive's, with a different meaningbin/loop_claude/routing.py:28first_route: the parameter anchor means the project's rung, re-glossed in nearly every docstringbin/model_routing/ladder.py:19Rung.__str__: parsed back to resume routing, so a wording change silently re-routes chunksbin/loop_claude/cycle/prompts.py:25routed_by: the routing policy stated a third time, beside main.py's help and routing.py's logic
2026-10-16T02:11:33Z, PR #283 at 7570af6, against 9aa0d4b
Ruling: worse
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | same | worse |
| Inherited architect | same | worse | worse | same | worse |
Mid A, decided by:
- Locality, Overall:
bin/rank_todo.py:502score: the score depends on the table's shape, branched on a string throughgoal_term, with the same old-versus-new check innotes,read_scoringandparse_item - Locality:
bin/rank_todo.py:94Scoring:weightsmeans different things by shape, so a reader ofscoring.weightsmust checkscoring.shapefirst - Locality:
bin/todo_table.py:17Moves: an old-shape Goals cell is stored as moves with distance None, soscorefalls back with(by or 0), andparse_citedtakes an unusedid_
Inherited architect, decided by:
- Locality:
bin/rank_todo.py:502score: reading one score needs the shape, decided elsewhere;weightsholds two things by shape and(by or 0)holds only becauseparse_citedreturns None moves - Locality, Shape:
bin/rank_todo.py:301unknown_columns: the old-shape path is checked at six sites acrossrank_todo.pywith its own reader and writer intodo_table.py, and nothing ties them together for whoever retires it - Shape:
bin/todo_table.py:16Moves: the name lies for old-shape rows, and the goals cell is parsed by a callable stored on a dataclass - Locality:
bin/todo_table.py:41parse_moves: the rules for a Moves cell are split betweenparse_movesandrank_todo.py'sparse_item
2026-10-16T05:16:12Z, PR #295 at 7ea23b1, against 1bff4ac
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | better | better | better |
| Inherited architect | better | same | better | same | better |
Mid A, decided by:
- Shape, Overall:
skills/larv-review/SKILL.md:36Principles: the principles audit is a pass of its own, with its own row and trigger, so the exceptions the base needed for it are gone - Self-sufficiency, Shape:
bin/model_routing/ladder.py:87review_rung: the rule that a reviewer runs on the higher rung is named, where the base wrotehighestinline twice - Self-sufficiency:
bin/model_routing/ladder.py:42Rung.parse: the spelling of a rung and its parser now sit beside__str__, where the base repeated the pattern inroute.py - Locality:
bin/rule_hook/reviewer_rung.py:11REVIEWERS: a list read from the disk at import, a hidden dependency that pulls towards worse
Inherited architect, decided by:
- Shape, Overall:
skills/larv-review/SKILL.md:36Principles: a pass of its own, with none of the base's exceptions to carry - Navigation:
docs/decisions.md:239routing: a chunk's worker runs on the rung its item picks…: the routing entry is split into titled entries, so a symptom leads to a heading - Shape:
bin/model_routing/ladder.py:87review_rung: one named rule that the worker's prompt and the hook share - Locality:
bin/rule_hook/reviewer_rung.py:13SEAT: the hook recognises a launch by a sentence in two markdown files, so rewording either turns the check off without a failure
2026-10-16T16:33:58Z, PR #296 at 50b9daa, against d1b0fe3
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | better | same | same |
| Inherited architect | same | worse | better | same | same |
Mid A, decided by:
- Locality:
bin/loop_claude/loop/reload.py:38decoded_as: looks up aRETIREDattribute and a staticfitsby name on whatever class it decodes, beside a second sentinel, so a rule on one class steers a generic module and only a comment says so - Locality:
bin/model_routing/route.py:11route_from: two parsers of one archived line, andRoute.__post_init__trusts that callers ranlevels_fitfirst - Locality:
bin/rule_hook/reviewer_rung.py:25reviewer_breaches: the required rung comes from a third environment variable, with a fallback to the old formula for a worker an older loop-claude started - Locality:
bin/rank_todo.py:570parse_item: a function named parse also computes R and writes it into the item's scores - Shape:
bin/model_routing/ladder.py:62worker_rung: worker, planning and review rungs are named for whom they serve, and plain step dicts replace the scan over band pairs - Shape, Self-sufficiency:
bin/loop_claude/cycle/prompts.py:26review_launches: one mapping gives each review's rung, and the docstring shows the sentence it renders - Shape:
bin/todo_table.py:23Shape: grew from 4 fields to 7, andscore_outcomes/outcomes.py'sreportdoes two jobs
Inherited architect, decided by:
- Locality:
bin/loop_claude/loop/reload.py:43decoded_as: a second sentinel and two class hooks found by name throughgetattr, so a dropped route at 3am depends on rules spread across two files and held by convention - Locality:
bin/model_routing/ladder.py:73worker_rung: raises KeyError for an L off its scale, safe only through guards injudge.py'sverdict_fromandRoute.__post_init__ - Locality:
bin/rule_hook/reviewer_rung.py:26reviewer_breaches: the reviewers' rung is worked out in loop-claude and reaches the hook through a third variable, beside an inline fallback restating the old rule - Shape:
bin/loop_claude/cycle/prompts.py:26review_launches:review_rungsis the one source the prompts, the worker's environment and the hook read, and the docstring shows the sentence produced - Shape:
bin/model_routing/ladder.py:63WORKER_STEPS: two small dicts with intent-revealing names replace the band tuple, andlowest_worker_rungreplaces a magic constant - Shape:
bin/todo_table.py:21Shape: namesoptional,risk_columnsandderivedas fields whereCwas hard-coded, though its seven-argument constructors take counting - Shape:
bin/model_routing/route.py:16ArchivedWorker: the file doubles, almost all of it for old R + C archive lines - Self-sufficiency:
bin/score_outcomes/outcomes.py:37SHOWN: its comment says C shows on any item scored before L, E and I, though only a table with A, G and Goals keeps it
2026-10-16T18:11:57Z, PR #298 at 3f8f30a, against 2c67b4f
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | same | same | same | same | same |
| Maintainability senior | same | same | same | same | same |
| Junior A | same | better | same | same | better |
Mid A, decided by:
- Locality, Overall:
bin/score_outcomes/main.py:86report_text: main only fetches the inputs and sets the exit code, and the report's sections are built apart from the git and fetch code
Junior A, decided by:
- Locality, Overall:
bin/score_outcomes/main.py:86report_text: the report is one pure function that states its section order and no longer writes R into the caller's dict, so one section changes without tracing the print order through main
2026-10-16T22:13:23Z, PR #301 at 2057d8f, against 11b9ec3
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | better | same |
| Inherited architect | better | worse | worse | better | same |
Mid A, decided by:
- Shape:
bin/model_routing/route.py:69Route.taking_in: sits besiderouting.taken_in, a near-identical name on another axis, a Rung picked by L against a Route holding an E + I, andafter_verdictchains both - Shape:
bin/loop_claude/cycle/judge.py:59Reply:bundled_escape_and_impactbesidebundled, which means the bundled L, so one name of each pair spells its axis and the other hides it - Shape:
bin/model_routing/route.py:12ROUTED: two small regexes became one dense one with optional levels, a retired R + C branch and a taken-in suffix, thoughroute_fromreads simpler - Self-sufficiency:
bin/loop_claude/routing.py:84judge_rung: the rule now lives inRoute.reviewed_on, which every consumer calls, in place of a pointer to a plan item - Self-sufficiency:
workflow-bench/review_git.py:1module docstring: states in place what it allows and why, where the bench skipped every permission check with no explanation
Inherited architect, decided by:
- Locality, Shape:
bin/loop_claude/loop/main.py:227after_verdict: one taken-in item updates the route in two places,steppedthroughtaken_inand the chainedtaking_in, with "bundled" a third word for the same idea - Locality:
bin/model_routing/route.py:73Route.reviewed_on:escape_and_impactno longer picks the reviewers, so a caller reading the field in place ofreviewed_on()gets a quietly low rung - Shape, Locality:
bin/model_routing/route.py:12ROUTED: one dense regex with nested optional groups, andlevels_fitaccepts more taken-in E + I thantaking_inkeeps - Shape:
bin/loop_claude/loop/main.py:231archive_step: a "Stepped up" outcome is written when only the reviewers' E + I rose, so the outcome's name no longer tells what happened - Navigation, Self-sufficiency:
bin/loop_claude/cycle/run.py:122keep_rejection: a loop that exits 1 after a rejected judge reply now leads to the archived reply, named likekeep_answer - Self-sufficiency, Navigation:
workflow-bench/review_git.py:1module docstring: explains the rule syntax and what the wrapper checks inline - Self-sufficiency:
bin/loop_claude/routing.py:84judge_rung: points toreviewed_onin place of a plan item
2026-10-17T01:10:16Z, PR #302 at f566d1c, against 46b79e7
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | worse | same | same |
| Inherited architect | same | worse | worse | same | worse |
| Maintainability senior | better | same | same | same | same |
| Junior A | same | worse | worse | same | worse |
Mid A, decided by:
- Navigation:
bin/loop_claude/loop/main.py:26__doc__: loop-claude's help namesbin/chunk-preflight, says it runs before routing and points toprompts.preflight_facts, so the worker's starting facts lead straight to the file and function - Navigation:
bin/todo_table.py:229chunk_item: the next chunk's item is picked in one named function both routing and chunk-preflight use, where46b79e7wrote it out insiderouted_item - Navigation:
bin/chunk_preflight/main.py:94preflight: the tool splits into main (effects), gathered (data and pure choices) and facts (wording), so a wrong printed line leads tofacts.pyand a wrong worktree choice togathered.worktree_add - Shape:
bin/todo_table.py:212standing: the items-table parser now also holds release policy (LastCut, CHANGELOG counting, a five-state standing, bundling), so its name no longer tells the truth, and themin(... named | {earliest} ...)test is the densest reading in the change - Shape:
bin/git_repo.py:74default_plan: the git wrapper imports the plan model and builds aLastCut, so a low-level module depends on a higher-level one
Inherited architect, decided by:
- Locality:
bin/loop_claude/loop/main.py:217run_loop: routing must run after the chunk-preflight subprocess, reading the plan as its fetch left it; only a comment guards that order, and a resumed session skips the preflight, where46b79e7read the working tree's todo.md in place - Locality:
bin/chunk_preflight/main.py:94preflight: some git calls take the checkout explicitly and others rely on the process's working directory, so seeing that they hit one repo needs how loop-claude launches the tool - Locality, Shape:
bin/git_repo.py:65DefaultPlan: the git plumbing module now knows todo.md and CHANGELOG.md and imports the plan model, so its name no longer says what it holds - Shape, Locality:
bin/todo_table.py:212standing: the table parser has taken on release policy that depends on todo.md texts back to the last tag; picking the routed item now runs through four modules, and the key test only makes sense beside the three-lineStandingcomment - Locality:
bin/loop_claude/cycle/run.py:28Cycle.facts: the first prompt names the chunk's item twice, from routing and from the preflight's facts, which agree only because both callchunk_itemand nothing lands on origin between the two reads - Shape:
bin/chunk_preflight/gathered.py:1gathered: it mixes dataclasses with plan, worktree, forge and path decisions, andfacts.pyis the renderer, so two names say less than they hold
Maintainability senior, decided by:
- Navigation:
bin/loop_claude/loop/main.py:26__doc__: the new paragraph says the preflight runs before routing and namesbin/chunk_preflight/andprompts.preflight_facts, and the routing paragraph namestodo_table.chunk_item, so why the worker was told X or routed to item N lands on the right file from --help - Navigation:
README.md:85"Tools": the new tool gets one line pointing to its --help, and the import paragraph says loop_claude runs it as a separate process - Navigation:
bin/todo_table.py:229chunk_item: one function picks for routing and the preflight, where 46b79e7'srouted_itemheld its own copy, so a wrong-item symptom leads to one unit
Junior A, decided by:
- Locality:
bin/loop_claude/loop/main.py:217run_loop: routing works only after the preflight, sincefetched_plannever fetches and counts on the preflight's fetch, which only a comment says; at46b79e7plan_itemread the working tree's todo.md wherever it ran - Locality, Shape:
bin/git_repo.py:65DefaultPlan: the git module importsLastCutfromtodo_tableand builds the plan pick's inputs, so its name and docstring no longer cover what it holds - Shape, Locality:
bin/todo_table.py:199landed: it counts ## Unreleased entries first whilelarv_review/git.py'slanded_since_releasecounts commits only, two meanings of "work has landed" - Shape, Locality:
bin/todo_table.py:212standing: the table parser grew from 247 to 346 lines and now decides how a release stands through a five-valued literal whose three-line comment must be decoded first, where 46b79e7'srouted_itemstated the pick in six lines - Locality:
bin/chunk_preflight/main.py:30__doc__: the preflight runs reading-cost in the new worktree, and larv-review's step now depends on whether the prompt shows that, so one action is split across a tool, a prompt and a skill - Shape:
bin/chunk_preflight/gathered.py:1gathered: the name covers the data but not the pure choices it also holds, andmain.pyimports 12 names from it on one line
2026-10-17T19:35:31Z, PR #303 at 85cbd6b, against 712dabd
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | same | same | same |
| Inherited architect | same | worse | same | same | same |
Inherited architect, decided by:
- Locality:
bin/comprehension_history.py:106latest_places: onemaxover the entries became three dependent steps, and thererun atrule in SKILL.md is needed to see why the start point moved - Locality:
skills/larv-review/SKILL.md:118"Second run" andbin/larv_review/plan.py:22scoring_run_due: the second run keys on the exact wording of a reason string, with no test or import linking the two - Locality:
bin/larv_review/agreement.py:10overall_cells: the last cell means overall only throughtable_rows' first-row-wins rule and the pre-#232Votetable, described atcomprehension_history.py:82 - Locality:
bin/larv_review/main.py:81__doc__: the help hard-codes 10 pairs whereMIN_PAIRSholds the number
2026-10-17T22:03:15Z, PR #308 at 023380f, against 59344de
Ruling: worse
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | same | worse |
| Inherited architect | same | worse | worse | same | worse |
Mid A, decided by:
- Locality, Overall:
bin/loop_claude/cycle/run.py:43Cycle.stepped_fromandescalated: the cycle carries route state outside its state machine, whichescalatedrewrites before every transition, keepingstepped_fromonly throughHumanTyped, soask,keep_answerandworker_starttrust the event before them - Locality:
bin/loop_claude/cycle/run.py:156unspent: whether a verdict keeps its escalation depends onescalation_spent, set by an earlier ruling of the cycle; escalation now spansescalated,unspentandloop/main.py's step fromconcluded.route - Locality:
bin/loop_claude/cycle/prompts.py:132typed/typed_answer: the typed answer is written into a prompt and parsed back out of the transcript, andrendertakes a callback the genericclaude_codemodule cannot make sense of alone
Inherited architect, decided by:
- Locality, Overall:
bin/loop_claude/cycle/run.py:71escalated: the route now also changes inside a cycle, through two fields kept out of the state machine; "why did the worker resume on this rung" spansescalated,unspent,routing.stepped_by_ruling,after_verdictand two archive-the-step sites - Locality:
bin/loop_claude/cycle/prompts.py:138TYPED_HERE/typed_answer: the resume prompt doubles as a data channel back to the judge, resting ontyped's layout, the blank-line partition andput_questionreturning one line, in three modules linked only by a comment - Shape:
bin/loop_claude/routing.py:80stepped_from: returns the stepped route, whileCycle.stepped_fromholds the route before the step;escalatedis likewise a function, a bool parameter and a local string - Shape:
bin/loop_claude/cycle/judge.py:95Interrupt: sharesescalationandbundledwithUnfinishedin the opposite positional order, and both are built positionally
2026-10-18T07:20:58Z, PR #310 at 1d8dd33, against fdd48bb
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | same | same | same |
| Inherited architect | same | same | same | same | same |
Mid A, decided by:
- Navigation:
bin/review_folder.py:5REPORTS: the review folder's path is one constant in a module named for it, imported bymake_review_folder,report_breachesandreviewed_in, so where reports go and who enforces it is found from one name
Inherited architect, decided by:
- Locality:
bin/todo_table.py:216renamed: one function now holds the renamed-never-finished rule thatnamed_releasesandplan.py'semptiedboth call, where a comment kept two files in step by hand - Locality:
bin/todo_table.py:205named_releases: now a fold whose answer depends on the texts' order, fixed far away by--reverseingit_repo.py'stodos_sinceand enforced by nothing - Locality, Self-sufficiency:
bin/stability_replay/main.py:183reviewed_in: three chained replaces that must run in a fixed order, and the copy placed under REPORTS for a reason that lives only inreviewer_report.py - Shape:
bin/stability_replay/main.py:175reviewed_in: seven positional parameters withcloneandreportsadjacent strings, and REPORTS meaning the root, its realpath and one round's copy in three files - Locality:
bin/review_folder.py:5REPORTS: the path is spelled again insettings.json, SKILL.md, CLAUDE.md and larv-review's help, and only the Python reads the constant
2026-10-18T07:59:13Z, PR #312 at 6694d54, against e074b8b
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same (voted better, no place naming Navigation) | better | same | same | better |
| Inherited architect | better | better | same | better | better |
Mid A, decided by:
- Locality, Overall:
bin/todo_table.py:223renames: one explicit## Renamedline in one text replaces a rename inferred from two plan snapshots - Locality:
bin/todo_table.py:209named_releases: an unparseable mark fails the result only in the newest text, an order dependence to keep in mind
Inherited architect, decided by:
- Locality, Overall:
bin/todo_table.py:223renames: the answer sits in the file read, whererenamedreplayed the plan's history - Locality:
bin/larv_review/plan.py:27emptied: renamed is a parameter, no hidden second derivation - Navigation, Self-sufficiency:
bin/chunk_preflight/gathered.py:100planned: the stop names the fix and where it goes
2026-10-18T09:10:02Z, PR #316 at 3eb3217, against 6e0fda4
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Locality, Shape:
bin/chunk_pick.py:92chunk_item: returns one of five dataclasses, each carrying its release, where 6e0fda4's(Standing, Item | None)tuple madeplannedrecompute the earliest release; the pick lives in one module named for it - Shape, Navigation:
bin/git_repo.py:1git_repo: the git wrapper no longer imports the plan model or builds todo.md objects;default_planandtodos_sincesit beside the logic that reads them - Shape, Locality:
bin/chunk_pick.py:97standing: namesfirst_namedand states the rule plainly, where 6e0fda4'smin(... named | {earliest} ...)had to be decoded - Shape:
bin/chunk_preflight/choices.py:1choices: the pure decisions leftgathered.py, so each module's name tells what it holds - Locality, Navigation:
bin/todo_table.py:269version: one release formatter besiderelease_of, where6e0fda4had three - Shape:
bin/chunk_pick.py:118unreleased_changes: withcommitted_since_release, the names say what is counted, where "landed" hid it
Inherited architect, decided by:
- Navigation, Locality:
bin/chunk_pick.py:97standing: "preflight says a release is ready to cut and it isn't" now lands in one module, where6e0fda4spread it over a table parser, a git wrapper importing it, and the preflight's wording - Locality, Shape:
bin/chunk_preflight/choices.py:24planned: each pick variant carries its release, so the messages read straight off it, where6e0fda4paired a recomputedearliestwith a string by hand - Shape:
bin/chunk_preflight/gathered.py:1gathered: data only, with the choices inchoices.py, thoughfacts.pyimportsPULLS_LISTfrom the decision module - Shape:
bin/chunk_pick.py:118unreleased_changes: withcommitted_since_releaseand oneversion, the names say what they count
2026-10-18T12:16:18Z, PR #321 at 862137b, against 0462980
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | better | better | same |
| Inherited architect | same | worse | better | better | better |
| Maintainability senior | same | better | better | better | better |
| Junior A | same | worse | same | better | same |
Mid A, decided by:
- Locality:
bin/loop_claude/cycle/judge.py:153escalation_mark: "one trigger buys one step" was one flag and one function inrun.py'sunspent; now a mark across cycles read throughRoute.escalated_at,entries_readstamped late byjudge_once, None versus 0,STEP_LINE,Grounds.since_stepand the routing steps, over three modules - Locality:
bin/loop_claude/cycle/judge.py:216quotes_checked: turns aQuotedAnswerinto anInterruptwithoutentries_read, correct only becausejudge_oncestamps afterrun_judge, which nothing here says - Locality:
bin/model_routing/route.py:12ROUTED: the mark survives only by round-tripping throughRoute.__str__, so rewording one fragment resets it to 0 at a distance - Shape, Self-sufficiency:
bin/model_routing/route.py:13ROUTED: a verbose regex with each group naming the__str__fragment it reads, in place of one dense line - Shape:
bin/loop_claude/cycle/judge.py:143Grounds: one named tuple whose fields document their own checks, in place of loose strings and a docstring list - Shape:
bin/loop_claude/routing.py:84one_rung_up: no longer shares its name withCycle.stepped_from, which means the route before the step - Self-sufficiency:
bin/loop_claude/cycle/prompts.py:87escalation: the docstring shows the full sentences it renders for a concrete rung
Inherited architect, decided by:
- Locality:
bin/loop_claude/cycle/judge.py:185judge_once: "one trigger buys one step" runs through five units in four modules and outlives the run via the archive line, with three "nothing" values for one idea: 0 for no mark, None for no escalation, None for never stamped - Locality:
bin/loop_claude/cycle/judge.py:219quotes_checked: builds anInterruptwithoutentries_read, safe only because the stamp comes later; swapped, the escalation is silently dropped - Shape, Self-sufficiency:
bin/loop_claude/cycle/judge.py:143Grounds: each checked text is a named field with its rule and penalty beside it, so the judge-rejection path reads easier - Shape:
bin/loop_claude/routing.py:84one_rung_up: the name says what it does, and itsescalated_atparameter shows it moves the mark, wherestepped_fromclashed withCycle.stepped_from - Locality, Shape:
bin/loop_claude/cycle/run.py:69escalated:Cyclelost the cycle-scopedescalation_spentandruleno longer filters throughunspent, sorun.pyreads on its own - Self-sufficiency:
bin/model_routing/route.py:13ROUTED: each group labelled by the fragment it reads, with a warning that a reword resets the mark - Self-sufficiency:
bin/loop_claude/cycle/prompts.py:88escalation: the docstring shows the rendered text for one concrete route - Self-sufficiency:
bin/loop_claude/cycle/judge.py:61STEP_LINE: defined beside the slicing that inserts it, with the entries-only-grow assumption written onRoute.escalated_at
Maintainability senior, decided by:
- Locality:
bin/model_routing/route.py:66Route.escalated_at: one persisted mark that holds across cycles replaces a cycle-scoped flag and a filter inrulethat rewrote the judge's verdict outside the judge - Locality:
bin/loop_claude/routing.py:76stepped_by_ruling: "no escalation, no step" now lives in routing, sorun.py'sescalatedreads straighter; offset in part byjudge_oncestampingentries_readafterquotes_checkedrebuilds anInterrupt, an order that matters - Self-sufficiency:
bin/loop_claude/cycle/judge.py:143Grounds: each checked text a named field with its own comment, in place of two positional strings and a docstring list - Self-sufficiency:
bin/model_routing/route.py:13ROUTED: each group commented with theRoute.__str__fragment it reads back - Self-sufficiency:
bin/loop_claude/cycle/prompts.py:87escalation: the docstring quotes the sentences it produces, and the body joins whole parts in place of space-led fragments - Shape:
bin/loop_claude/routing.py:84one_rung_up: renamed fromstepped_from, which meant the opposite ofCycle.stepped_frombeside it
Junior A, decided by:
- Locality:
bin/model_routing/route.py:66Route.escalated_at: "one trigger buys one step" moved from one file's cycle-local flag andunspentfilter to a mark threaded through six files - Locality:
bin/loop_claude/cycle/judge.py:189judge_once: stampsentries_readafter the run, an unstamped None silently steps nothing, and "none" is 0 on the route but None insteppedandstepped_by_ruling - Self-sufficiency:
bin/model_routing/route.py:13ROUTED: a verbose pattern whose groups name the fragments they read, in place of one dense line - Self-sufficiency, Shape:
bin/loop_claude/cycle/judge.py:143Grounds: each field documents what must quote it and what a misquote does, thoughjudgeandjudge_oncegrow to nine positional parameters - Shape:
bin/loop_claude/routing.py:84one_rung_up: removes the name it shared withCycle.stepped_from, which meant the opposite - Self-sufficiency:
bin/loop_claude/cycle/prompts.py:87escalation: a worked example of the exact sentences, as its neighbours already had
2026-10-18T16:57:38Z, PR #322 at 49e86fd, against 2911d3f
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | better | same | better |
| Inherited architect | same | worse | better | same | same |
| Maintainability senior | same | better | worse | same | same |
| Junior A | better | better | same | same | better |
Mid A, decided by:
- Navigation, Shape:
bin/review_replay/main.py:1review_replay: the tool takes any pass and its name says so, withREVIEWERSlisting the passes and the usage printing them, where the old name pointed at stability alone - Shape:
bin/review_replay/main.py:208reviewed_in: copy, rewrite, clone and run are each a named step with a docstring, wherebin/stability_replay/main.py:175did all four in one body;codes_foundtakes the left-out loop out ofmain - Locality:
bin/rule_hook/reviewer_rung.py:18reviewer_breaches: asksreview_rungs(...)[kind]directly, without the older loop-claude fallback or the filesystem glob - Locality:
bin/loop_claude/cycle/prompts.py:72taken_in_launches: compares wholereview_rungsdicts, right only because of a condition insideladder.review_rungsthat a comment states - Locality:
bin/model_routing/ladder.py:71REPLAYED: an empty table thatreview_rungs,lowered,shared_launchesandrerunsall branch on, so a reader follows a path no data drives - Shape:
bin/rule_hook/reviewer_report.py:10REVIEWERS: a frozenset of agent files, whileladder.py'sREVIEWERSis a dict from pass to agent; one name for two shapes - Shape:
bin/loop_claude/cycle/prompts.py:43review_launches: fan-out grows by four units and "each reviewer" becomes a magic key, but stays bounded with docstring examples
Inherited architect, decided by:
- Locality:
bin/loop_claude/cycle/prompts.py:43review_launches: one launch sentence now needsREVIEWS' per-pass rows,shared_launchesskipping them for an"each reviewer"key, andloweredputting them back, where the base was one loop over three rows - Locality:
bin/loop_claude/cycle/prompts.py:78taken_in_launches: compares the wholereview_rungsdict, correct only by a rule enforced inladder.review_rungs - Locality:
bin/model_routing/ladder.py:71REPLAYED: a hand-edited table, empty today, that overrides a pass's rung below the worker's under a combined E + I and worker condition; a second source of truth - Shape:
bin/review_replay/main.py:208reviewed_in: split intocopy_reports,redirectedandclone_reviewed, each with a docstring;codes_foundandpass_and_modeltake inline loops and lambdas out ofmain - Shape:
bin/rung_replay/command.py:16arguments_or_exit: argument parsing is no longer hard-wired to one model, anddown_from(top, floor)makes judge-replay's floor show at its call site - Shape:
bin/rule_hook/reviewer_report.py:10REVIEWERS: shares its name withladder.py's hand-kept dict of a different shape
Maintainability senior, decided by:
- Shape:
bin/loop_claude/cycle/prompts.py:14REVIEWS: six rows built fromREVIEWERS, a separate"each reviewer"entry, akind not in REVIEWERSfilter and aloweredlist re-joined in two places, all serving an emptyREPLAYED - Shape:
bin/model_routing/ladder.py:69REVIEWERS: two modules on one rule-hook call path each hold aREVIEWERSof a different type and meaning, where the base had one - Shape:
bin/model_routing/ladder.py:104review_rungs: fans out to every pass then overlaysREPLAYEDunder a condition only a dense docstring paragraph explains;taken_in_launchescompares whole dicts - Locality:
bin/review_replay/main.py:208reviewed_in: copy, rewrite and clone are named units, and the lambdas passed toreplayedare named functions - Locality:
bin/rung_replay/command.py:35down_from: takes the floor as an argument, so each caller states its own floor where the base hid judge-replay's policy in the shared helper
Junior A, decided by:
- Locality:
bin/rule_hook/reviewer_rung.py:24reviewer_breaches: the required rung comes straight fromreview_rungs, the function the prompts use, where the base passed it through an environment variable and a fallback - Locality:
bin/rule_hook/reviewer_report.py:10REVIEWERS: report confinement readsagents/itself, no longer importing its set from a module about rungs - Locality, Shape:
bin/review_replay/main.py:208reviewed_in: four jobs and seven arguments became named units with docstrings, andmain's loop and lambdas are named - Navigation:
README.md:113: the import map now names every reader ofmodel_routing/and whatrung_replay/holds, and the tool's name leads to it for any pass - Shape:
bin/loop_claude/cycle/prompts.py:57shared_launches: five helpers, a made-up"each reviewer"key and paths an emptyREPLAYEDnever runs, againstReviewKindnaming the real passes
2026-10-18T19:59:26Z, PR #324 at c713910, against e6223f7
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | better | same | same |
| Inherited architect | same | worse | better | same | same |
Mid A, decided by:
- Locality:
bin/larv_review/record.py:56rerank: recording a scoring run now also rewrites the checkout's todo.md and can exit 1 on items under the bar, so changing record-scoring needs rank_todo.rank_file's whole contract, where the base only appended to the history - Locality:
bin/rank_todo.py:338weigh: an item's score now depends on docs/comprehension-history.md's newest scored commit and on a## review opt-out:title in docs/decisions.md, so scores and --check can change without anyone editing todo.md or README.md - Shape:
bin/rank_todo.py:315ranked_goals: the base's inline three-way weights branch in read_scoring is a named weigh, with the code-health and void rules in ranked_goals, each easier alone, at the cost of three similar names weigh, goal_weights and parse_weights - Shape:
bin/rank_todo.py:623cited_goals_failure: the base's parse_item mixed goal checks with writing R, and main inlined read, rank, write and print; each is now a unit with one purpose, and rank_file is reusable - Shape:
bin/comprehension_history.py:130newest_scored: the base's index arithmetic in latest_places is a named helper latest_overall shares, and overall_cells and FLOOR moved to the module that parses the table they read
Inherited architect, decided by:
- Locality:
bin/rank_todo.py:196read_all: ranking now also depends on the newest scoring run in docs/comprehension-history.md, so a passing --check can fail with nobody touching todo.md, and nothing at the call site shows that a history append moves code health's weight - Locality:
bin/larv_review/record.py:56rerank: recording a scoring run now writes todo.md through rank_todo.rank_file, which prints its notes to stderr, and the message pointing at "the lines above" reads right only if rank_file printed first - Locality:
bin/rank_todo.py:106OPTED_OUT: ranking now matches a docs/decisions.md title spelled in larv-review's SKILL.md → "Opt-out", so rewording that heading silently changes whether rank_todo refuses a plan - Locality:
bin/loop_claude/cycle/prompts.py:17AFTER_CODE_HEALTH: the caveat that code health ranks first unless the README voids it is repeated in prose across about eight files, so one change to the shared goal edits all of them - Shape:
bin/comprehension_history.py:130newest_scored: replaces the base's index juggling in latest_places with two truthfully named predicates serving latest_places and latest_overall - Shape:
bin/comprehension_history.py:145overall_cells: FLOOR and overall_cells now have one home next to DIMENSIONS, and agreement's inline sort became the named scoring_runs - Shape:
bin/rank_todo.py:332weigh: weigh, cited_goals_failure and rank_file are pulled out of read_scoring, parse_item and main, each doing one thing
2026-10-18T19:59:40Z, PR #324 at c713910
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6.5 | 6 |
| Mid A | 7 | 5.5 | 6 | 6.5 | 6 |
| Maintainability senior | 7 | 6 | 6.5 | 6.5 | 6.5 |
| Inherited architect | 7 | 6 | 6 | 7 | 6.5 |
| Mean | 7.00 | 5.75 | 6.13 | 6.63 | 6.25 |
Junior A, hardest first:
bin/model_routing/route.py:66Route.escalated_at: the escalation mark lives in six units across four files, with judge_once stamping entries_read, escalation_mark, stepped and judge_prompt's STEP_LINE held at oncebin/loop_claude/loop/main.py:1"loop-claude --help": a 164-line docstring is the loop's real spec, its routing paragraphs packing every rule into one block that routed and after_verdict need openbin/loop_claude/cycle/prompts.py:46review_launches: prompt sentences assembled from rung comparisons over replay-lowered passes, the project's rung, the top rung and higher E + Ibin/model_routing/route.py:13ROUTED: stored state persisted as prose and parsed back by regexes in four places, so rewording one string silently changes routingbin/rank_todo.py:568score: a 660-line file carrying two table shapes, with OLD_SHAPE forks in scoring, weights, goal parsing and column checksbin/chunk_pick.py:97standing: five inputs become five outcomes, and chunk_pick.Finished shares its name with judge.Finishedbin/review_replay/rounds.py:73reviewed_path: rebuilds the code a past reviewer read from prompt text and author times, a heuristicbin/comprehension_history.py:159scored: legacy formats from before PR #255 and #232 leak into the reading
Mid A, hardest first:
bin/loop_claude/routing.py:64stepped: seven small functions, each defined through another, with six inputs held at once to predict one rung, and an import cycle shaping the signaturebin/loop_claude/loop/main.py:259after_verdict: the 164-line docstring is the only full map of routing, and after_verdict chains it all in one expressionbin/loop_claude/cycle/run.py:45Cycle.stepped_from: the route changes outside the state machine, cleared on every event but HumanTyped, so it outlives exactly one hopbin/loop_claude/cycle/judge.py:216quotes_checked: three bodies of text to check quotes against, fuzzy matching, and three ways a failed check is handledbin/loop_claude/cycle/prompts.py:46review_launches: fragments that each start with a space, so no whole prompt is visible without running it in your headbin/model_routing/route.py:13ROUTED: a saved string format parsed back by a verbose regex that still accepts an older formatbin/rank_todo.py:568score: two table shapes share every path, the goal term W in one and V in the otherbin/loop_claude/loop/reload.py:43decoded_as: a generic decoder with two sentinels and a duck-typed fits hook defined on Route
Maintainability senior, hardest first:
bin/loop_claude/routing.py:64stepped: four things held at once, the unfinished count, the taken-in L, the escalation cap from item_rung and the mark threaded through judge.py's escalation_mark, STEP_LINE and Grounds.since_stepbin/model_routing/route.py:86Route.__str__: the route written as English and parsed back by ROUTED and outcomes.workers, so a sentence's wording is a storage schema shared by four toolsbin/loop_claude/loop/main.py:1"loop-claude --help": a 160-line spec whose routing paragraph carries about 15 rules, re-describing what routing.py, ladder.py and retry.py enforcebin/loop_claude/cycle/prompts.py:46review_launches: each sentence built conditionally with a leading-space convention, and taken_in_launches scanning every higher E + Ibin/loop_claude/cycle/judge.py:216quotes_checked: three grounds with three failure consequences, and the demotion to an interrupt loses release_readybin/review_replay/rounds.py:73reviewed_path: guesses which removed worktree a reviewer read, and redirected depends on the order of its replacementsbin/comprehension_history.py:150scoring_runs: near-synonyms scored and scoring_runs differ only on a ruling from before PR #255bin/rank_todo.py:332weigh: the goal term is W or V by shape, so every scoring path asks which shape
Inherited architect, would restructure first:
bin/round_titles.py:4CAP: no program loops the reviewers; the cap lives in SKILL.md prose and the code only counts rounds and prints a notebin/git_command.py:16commits: parses git commit messages for rule-hook while git_repo.py, rule_hook/git.py, larv_review/git.py and Effects.git each run git, and git_repo bypasses Effectsbin/transcript_archive/outcomes.py:39outcomes: shares its name with score_outcomes/outcomes.py, which measures finished chunks insteadbin/loop_claude/routing.py:64stepped: routing has four homes, ladder.py, route.py, routing.py and routing_check with routing_figures, and prompts.py restates the ladder as sentencesbin/rank_todo.py:196read_all: a 658-line library as well as a tool, imported by rule_hook and larv_review, and now reading four documents, so a todo.md commit in any project depends on its comprehension historybin/loop_claude/cycle/judge.py:68Reply.bundled: the taken-in L is also called bundled, and the project's rung is also called anchorbin/chunk_pick.py:97standing: "next release" is computed two ways, here and in larv_review/plan.py's scoring_run_due, which may diverge after a rename
2026-10-18T19:59:40Z, PR #324 rerun at c713910
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6.5 | 6 | 6 |
| Mid A | 7.5 | 6 | 6.5 | 7 | 6.5 |
| Maintainability senior | 7.5 | 6 | 6.5 | 6.5 | 6.5 |
| Inherited architect | 6.5 | 6 | 5.5 | 6.5 | 6 |
| Mean | 7.13 | 5.88 | 6.25 | 6.50 | 6.25 |
Junior A, hardest first:
bin/loop_claude/cycle/prompts.py:46review_launches: the worker's English instructions built from how the worker's, the project's, the reviewers' and any replay-lowered rung relate, with ladder.review_rungs, REPLAYED and escalation_rungs open at oncebin/loop_claude/routing.py:64stepped: seven small functions defined through each other, bundled learned only from judge.Reply, and escalated_at both an argument and a Route field with different meaningsbin/model_routing/route.py:13ROUTED: routing state lives as English sentences in outcomes.md, read back by a regex, with the newest file by mtime decidingbin/loop_claude/loop/main.py:1"loop-claude --help": about 160 lines and the only full account of routing, step-ups, the archive layout, reloads and the routing figures in one unbroken paragraphbin/loop_claude/cycle/run.py:69escalated: stepped_from must survive a HumanTyped event and reset on every other, and archive_stepped writes one outcome or two depending on itbin/loop_claude/cycle/judge.py:216quotes_checked: three quote checks follow different rules, and an escalation that fails is silently droppedbin/chunk_pick.py:97standing: two separate notions of the next release and of work having landed, here and in larv_review/plan.py's scoring_run_duebin/rank_todo.py:332weigh: old-shape support forks behaviour throughout a 659-line file
Mid A, hardest first:
bin/model_routing/route.py:66Route.escalated_at: the escalation mark is one concept spread over five files, the entry count, the injected step line, a double quote check and the cap from item_rung held at oncebin/loop_claude/cycle/run.py:69escalated: state outside the state machine, cleared on every event but HumanTyped and read by ask, keep_answer and archive_stepped, depending on its order before transitionbin/loop_claude/cycle/prompts.py:46review_launches: routing policy and English wording interleaved, much of the branching serving an empty REPLAYEDbin/loop_claude/loop/main.py:210run_loop: the route passed as lazy thunks and re-derived every cycle, with an anchor-is-None branch in three placesbin/model_routing/route.py:86Route.__str__: the archive's human-readable lines are what reruns, the routing figures and the replays parse backbin/comprehension_history.py:150scoring_runs: two near-identical predicates differing only for a ruling from before PR #255bin/rank_todo.py:568score: every rule exists twice for the old table shapebin/review_replay/main.py:149code_of: rebuilds which worktree a reviewer read by the first missing directory among its prompt's paths
Maintainability senior, hardest first:
bin/loop_claude/routing.py:64stepped: seven one-line rung functions each defined by the others, four inputs held together, and an import cycle steering the interfacebin/loop_claude/cycle/judge.py:153escalation_mark: the escalation mark spread over route.py, judge.py's Grounds and judge_prompt and run.py's escalated, and no single place states that one trigger buys one stepbin/loop_claude/cycle/prompts.py:46review_launches: which conditional sentences reach which prompt means tracing fresh, routed_by, resumed, rungs_held and stepped_tobin/loop_claude/loop/main.py:1"loop-claude --help": the routing paragraph is one block covering the ladder, the judge, replays, rule-hook's variables, step rules, archive lines and legacy R + Cbin/model_routing/route.py:13ROUTED: routing state survives restarts as English sentences parsed back by regex, a hidden contract across three packagesbin/rank_todo.py:218rank: every stage branches on the old table shape, about half a 660-line file read and then discardedbin/comprehension_history.py:130newest_scored: four near-identical names over two concepts, split by a legacy case only the docstrings explainbin/chunk_pick.py:97standing: five inputs, three of which may be Failure, produce five outcomes, the order of the Failure checks carrying meaning
Inherited architect, would restructure first:
bin/round_titles.py:5PASSES: the pass list lives in four places, PASSES, ladder.ReviewedPass, REVIEWERS and SKILL.md's table, kept in sync by handbin/rule_hook/rules.py:55round-title-counts: no program blocks a fourth review round; rule-hook checks only the counts inside a round's titlebin/git_repo.py:8git: bin/'s root holds about 43 entries, with 13 loose modules such as git_repo, git_command, chunk_pick and todo_table not grouped by purposebin/loop_claude/cycle/judge.py:68Reply.bundled: the taken-in L and E + I are bundled in the judge's reply and taken in on Route and in routing.taken_inbin/model_routing/route.py:86Route.__str__: the archive's text format is written in two modules and parsed back by ROUTED, routing_figures, reading_cost, score_outcomes and judge_replaybin/todo_table.py:89OLD_SHAPE: the old todo.md shape is the largest dead weight in rank_todo.py
2026-10-21T18:16:09Z, PR #333 at 7c7a398, against fda48ac
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Shape:
bin/git_repo.py:22git: now the one git runner, with a guarded flag and GUARD_SETTINGS giving each transport its own line and saying why, where four more wrappers each had their own failure text and decoding - Shape, Locality:
bin/loop_claude/effects.py:16Git: the effect returns str | Failure like everywhere else, removing checkout.py's try/except OSError and returncode handling - Shape:
bin/rule_hook/git.py:7git: a dense one-line GUARDS comprehension became a thin call with guarded=True whose docstring says why - Self-sufficiency:
bin/rule_hook/rules.py:64OVERRIDE: the new comment explains the lazy reason, the lookahead and the stripped --> - Self-sufficiency:
CLAUDE.md:108"Vocabulary": now defines judge's records, paired ruling, place, scoring run and seat - Self-sufficiency:
bin/metric_bench/main.py:114run_scores: a small pull the other way, the comment cites plan item 204
Inherited architect, decided by:
- Navigation, Shape, Overall:
bin/git_repo.py:22git: four places that started git with differently worded errors now go through one function, and the hardening is one named list instead of a one-line comprehension - Locality, Shape:
bin/loop_claude/effects.py:16Git: Effects.git returns str | Failure, so checkout.py and review_replay drop their own exception and returncode conventions - Locality:
bin/metric_bench/main.py:112run_scores: what the tool does on bad input is stated in its help, not hidden in a private wrapper's errors="replace" - Self-sufficiency:
CLAUDE.md:105"Vocabulary": defines the comprehension terms the charter used without defining, and rules.py's OVERRIDE gains its explanation - Self-sufficiency:
bin/metric_bench/main.py:114run_scores: a small cost, "item 204's" points to a plan item a code reader can't resolve - Locality:
bin/rule_hook/git.py:7git: a small cost, the guards no longer sit beside the hook that needs them, though both docstrings say why
2026-10-21T20:20:12Z, PR #335 at 33645b0, against dec19c5
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | better | same | better |
| Inherited architect | better | same | better | same | better |
Mid A, decided by:
- Navigation, Shape:
skills/larv-review/SKILL.md:73"Re-entry": one table now answers which passes a fix reruns, where five Trigger cells and three bullets did, with a worked example under it - Shape:
skills/larv-review/SKILL.md:86"What a round reviews": the round-2 scope left "Round cap", which now holds only counting and capping, as its name promises - Locality:
bin/settings_filter.py:12cleaned: hidden state, since the stored and the working file differ and a per-clone git setting turns the filter on, with an ordering rule for Update; it balances the Re-entry table's gain - Self-sufficiency:
bin/settings_filter.py:2module docstring: the tool explains itself, but the steps to run it live in README and AGENTS.md
Inherited architect, decided by:
- Navigation, Shape:
skills/larv-review/SKILL.md:73"Re-entry": "which passes does this fix rerun" is answered in one table where it was split between Trigger-cell clauses and a Rerun bullet - Shape:
skills/larv-review/SKILL.md:31pass table's Trigger column: each cell now says only when the pass first runs - Navigation:
skills/larv-review/SKILL.md:86"What a round reviews": a named paragraph beside Re-entry, where a reader asking what round 2 looked at searches - Locality:
.gitattributes:1the settings filter: committed behaviour depends on an untracked per-machine git setting, and README's Install and Update steps work only in order - Self-sufficiency:
bin/settings_filter.py:2module docstring: the script stands alone, but cannot say it is inert unless Install's git config line ran
2026-10-22T18:56:17Z, PR #354 at 7042a75, against 6f58a0e5c2
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | same | better | better |
| Inherited architect | better | better | same | better | better |
Mid A, decided by:
- Navigation, Locality:
bin/review_passes.py:7PASS_AGENTS: the pass names lived in two files, one of them the routing ladder where nobody would look; one module named for its contents holds both now - Navigation, Self-sufficiency:
bin/loop_claude/loop/main.py:53Routing: the worker's rung: the 70-line routing paragraph is now headed parts a reader jumps to - Navigation, Self-sufficiency:
bin/transcript_archive/__init__.py:1"transcript_archive": the archive's file layout sits with the code that writes it, in place of loop-claude's help - Locality:
bin/larv_review/plan.py:32landed_towards: the next release is asked ofchunk_pick, so the rule is written once - Self-sufficiency:
bin/shell_command.py:33segments: the new docstring lists each shell form it does not parse - Self-sufficiency:
bin/git_repo.py:3"git_repo": the module docstring says every tool starts git through it
Inherited architect, decided by:
- Navigation, Locality:
bin/review_passes.py:7PASS_AGENTS: two unlinked pass lists became one table in a module named for it, so adding a pass is a single edit - Locality:
bin/larv_review/plan.py:32landed_towards: it reuseschunk_pick's rule in place of its own, so the scoring-run trigger and chunk-preflight cannot disagree about the next release - Locality:
bin/larv_review/ruling.py:38surplus: the surplus-seat check is a named predicate called where the ruling is recorded, not hidden invotes - Navigation:
bin/loop_claude/loop/main.py:53Routing: the worker's rung: seven headed sections replace one 50-line routing paragraph - Self-sufficiency:
bin/shell_command.py:33segments: the docstring answers why rule-hook missed a git call - Self-sufficiency:
bin/rule_hook/override_log.py:36this_chunks: the case list maps each commit shape to its one-line return - Self-sufficiency:
bin/larv_review/main.py:34"larv-review --help": the help no longer states when work has landed, and points toplan.pyinstead - Locality:
bin/score_outcomes/main.py:134read_outcomes: whether a commit joins the chunk above depends on the previous iteration's state, with a positionalChunktuple - Locality:
bin/todo_table.py:137NoItemsTable: every caller must test it beforeFailure, and swapping the checks silently changes the message
2026-10-22T20:32:06Z, PR #357 at 3f8e175, against 64023b5
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Shape:
bin/loop_claude/routing.py:67after_unfinished: takes the verdict itself and owns the whole route after an unfinished verdict, wheresteppedtook loose ints, explained an import-cycle workaround and lefttaking_into the caller - Locality, Navigation:
bin/loop_claude/loop/main.py:251run_loop:routedreturns a Route andafter_verdicttakesConcluded, so the off-ladder None and theended/beforejuggling are gone andarchive_stepcompares two records in one pass - Navigation, Shape:
bin/loop_claude/verdicts.py:1"verdicts": the verdict types left the longcycle/judge.pyfor a small module whose docstring says why it sits outside cycle/ - Shape:
bin/loop_claude/cycle/judge.py:69Reply.highest_likelihood_taken_in:bundledgave no hint it carries an L; the new name does, and the verdicts are built by keyword - Navigation, Shape:
bin/chunk_pick.py:50ReadyToCut: no longer clashes with the judge'sFinished - Shape:
bin/chunk_pick.py:109next_release: replaces the five-argumentstandingand is shared with larv_review'slanded_towards;Upcoming.is_earliestneeds its comment - Locality:
bin/model_routing/route.py:74Route.record: a typed JSON record replaces the prose thatROUTEDparsed back, at the cost of a second, labelled read path
Inherited architect, decided by:
- Locality, Shape:
bin/loop_claude/routing.py:67after_unfinished: takes the Unfinished verdict and does all the next route's work, the E + I intake included, wheresteppedtook loose ints because of an import cycle - Locality:
bin/loop_claude/loop/main.py:274after_verdict: one call replaces the base's order-sensitiveendedcopy andbeforeunpacking, andarchive_stepnames both routes outright - Shape:
bin/loop_claude/cycle/judge.py:122verdict_from: verdicts built by keyword where positional calls met differing field orders, andbundledrenamed to what it holds - Navigation, Shape:
bin/chunk_pick.py:50ReadyToCut: ends the two-Finishedclash, andnext_releasesplits the five-argumentstanding - Self-sufficiency, Locality:
bin/model_routing/route.py:13Record: the archived worker line is a typed JSON record the outcome writers alone accept, with a warning that renaming a field orphans records; the legacy regex is unannotated and R + C lines fall throughrung_atas text without a docstring saying so - Navigation, Self-sufficiency:
bin/loop_claude/verdicts.py:1"verdicts": its own module, with field comments stating whatentries_read=Nonemeans - Self-sufficiency:
bin/todo_table.py:205ships_alone: the R above 5 rule is one named function, where the base held it only in the charter
2026-10-23T02:05:49Z, PR #359 at 72a63c5, against 545ab3a
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Shape, Overall:
bin/model_routing/review_choice.py:21chosen: the choice of each review's rung has one home, returned as a ReviewChoice; before, three functions in prompts.py each worked it out - Locality, Shape:
bin/loop_claude/cycle/prompts.py:62review_launches: the sentence builders return lists of sentences and one join adds the spaces, where each fragment had to start with a space - Shape:
bin/loop_claude/routing.py:331first_route: the parameter anchor is now called project, so docstrings no longer explain it - Navigation, Self-sufficiency:
bin/loop_claude/routing.py:1"module docstring": it lists the unit for each routing phase, where the base pointed to docs/decisions.md - Navigation:
bin/loop_claude/loop/main.py:54"loop-claude --help": about 90 lines of routing text are one paragraph that names the file for each concern - Self-sufficiency, Navigation:
bin/loop_claude/cycle/prompts.py:1"module docstring": new, says which function makes which sentence group and where each lands - Locality:
bin/model_routing/ladder.py:113replay_applies: the condition for when a replay applies is a named function shared by review_rungs and chosen
Inherited architect, decided by:
- Locality, Shape, Overall:
bin/loop_claude/cycle/prompts.py:55review_launches: the choice is made once in chosen and the prompt code turns it into lists of sentences; the base called review_rungs again in three functions - Shape:
bin/loop_claude/cycle/prompts.py:17EACH_REVIEWER: the REVIEWS table of wording and kind pairs, filtered in two places, is gone - Navigation, Self-sufficiency:
bin/loop_claude/routing.py:1"module docstring": a four-line map of Start, Rerun, Step-up and Judge naming the function for each, in place of about 90 lines of routing prose in the help - Navigation:
bin/loop_claude/loop/main.py:54"loop-claude --help": the routing paragraph names a file for each concern - Navigation:
bin/loop_claude/cycle/prompts.py:1"module docstring": says which function builds which part of each prompt - Shape, Locality:
bin/model_routing/ladder.py:113replay_applies: the replay condition has one name, shared by review_rungs and chosen - Shape:
bin/loop_claude/cycle/run.py:41Cycle.project: the rename from anchor removes the repeated explanation
2026-10-23T03:24:42Z, PR #361 at 6f98a18, against 39dd51b
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Shape:
bin/loop_claude/cycle/run.py:61run_cycle: the step-up rides on Ruled and the states as a StepUp passed to work, ask, keep_answer, worker_start and archive_stepped, where the base's Cycle.stepped_from was hidden state that escalated() cleared on every event but HumanTyped - Navigation, Self-sufficiency:
bin/loop_claude/escalation_mark.py:1escalation_mark: the mark's rule, once spread over judge.py, verdicts.py, Route.escalated_at's comment and after_unfinished's docstring, sits in one file whose docstring lists Stamp, Step, Show and Check by function - Shape:
bin/loop_claude/escalation_mark.py:37moved_to: says what it returns, where the base's verdicts.escalation_mark was named after the concept; Escalating replaces the repeated union - Shape, Locality:
bin/loop_claude/cycle/judge.py:164quotes_checked: shrinks to calls into escalation_mark, so judge.py holds only the judge's own logic, and verbatim and plain sit beside render - Self-sufficiency:
bin/round_titles.py:6CAP: a bare constant in the base now says what it caps and which two consumers use it
Inherited architect, decided by:
- Locality, Shape:
bin/loop_claude/cycle/run.py:61run_cycle: a StepUp travels through Ruled, Interrupting and Working and drops on /talk in one visible case, where the base's escalated() cleared Cycle.stepped_from by an order-dependent rule; the base's hardest place improved most - Navigation, Self-sufficiency:
bin/loop_claude/escalation_mark.py:1escalation_mark: from "the worker was stepped up twice" a reader reaches the right unit in one hop, with four labelled steps each naming its function - Shape:
bin/loop_claude/verdicts.py:50Escalating: the base's escalation_mark function returned where the mark would move; now the module carries the concept's name, moved_to says what it returns, and the type has one name - Self-sufficiency:
bin/loop_claude/cycle/judge.py:70Reply.escalation: the comment points at code that explains the rule rather than at docs/decisions.md - Navigation:
bin/round_titles.py:6CAP: names the decision, rule-hook's round-cap and larv-review round-count, so a fourth-round page leads to the code that enforces the cap
2026-10-23T09:16:16Z, PR #367 at 5a56fff040, against dcbdc970a6
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | same | better | same | better |
| Inherited architect | better | same | better | better | better |
Mid A, decided by:
- Navigation, Shape:
bin/larv_review/history_changes.py:6history_changes: named by its job wheregit.pynamed only that it calls git, and the merge-commit rule has its own unit inhistory_change - Navigation, Shape:
bin/larv_review/history_commands.py:100show_places: the module's name now fits all six entry points, whererecord.pyalso held commands that record nothing - Shape:
bin/model_routing/route.py:13WorkerLine:Record,Route.recordandRECORD_FIELDSclashed with three other meanings of record; the new names match theWorker:line - Shape:
bin/score_outcomes/results.py:24Result: no longer collides with the archive's outcomes, at the cost of a comment explaining why the tool still says outcome - Shape:
bin/judge_replay/replay.py:19archived_kind: says it holds the archived verdict kind, whereon_recordread like a flag
Inherited architect, decided by:
- Shape, Self-sufficiency:
bin/model_routing/route.py:13WorkerLine: says the value is the JSON on theWorker:line that ends each outcome, whereRecordoverlapped the document sense and the verb - Navigation, Shape:
bin/larv_review/history_changes.py:6history_changes: a ruling-base symptom now leads to a file named for its one job, and the first-parent rule sits in one small unit - Navigation:
bin/larv_review/history_commands.py:16record_ruling: imported from a file named for the history commands it serves, not the overloaded record - Shape:
bin/rule_hook/rules.py:91Checked: no longer reuses outcome and verdict, two words the project uses for archive entries and judge verdicts - Shape, Self-sufficiency:
bin/reading_cost/measure.py:13Reading:filesno longer borrows the seats' place, and the docstring says what both fields count - Shape:
bin/judge_replay/replay.py:24Case.archived_kind: says it is the kind the archive gave the verdict - Self-sufficiency:
bin/score_outcomes/results.py:22Result: a reader translates between result in the code and outcome in the tool's name and report, a small cost the comment explains
2026-10-23T09:51:07Z, PR #368 at 5e7e364, against 4a0ba59
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | better | same | better |
| Inherited architect | same | better | better | better | better |
Mid A, decided by:
- Locality, Shape:
bin/score_outcomes/results.py:40MEDIANS: readsresult.findings.redfrom the namedFindingstype where the base readresult.findings[0]and sent the reader toround_findingsto learn index 0 means 🔴 - Locality, Shape:
bin/score_outcomes/results.py:138at_least: the lower-bound rule and its reason now live in one named function thatrenderandmisreadsshare, withcountedbeside it - Shape, Locality:
bin/round_titles.py:45title_counts: one parse of a round title's counts where the base matchedFINDINGSseparately in each caller;lacks_countsnow also means inconsistent counts, a small cost
Inherited architect, decided by:
- Shape, Self-sufficiency:
bin/score_outcomes/results.py:33Result.findings: a namedFindingstype in place of a baretuple[int, int], so each use reads.redor.yellowwith no docstring open - Locality:
bin/round_titles.py:45title_counts: one function parses the counts after a round title's colon, so a change to the title format happens in one place - Locality:
bin/score_outcomes/results.py:143counted:renderandmisreadsshare one spelling of a chunk's findings, so the two output paths cannot drift apart - Shape:
bin/round_titles.py:34lacks_counts: the name now also covers a misread count above 🔴 plus 🟡, which a reader chasing a refusal learns only from the body; a cost that doesn't outweigh the gains
2026-10-23T09:51:56Z, PR #366 at e05c54a, against 4a0ba59
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | worse | same | same |
| Inherited architect | same | same | worse | same | same |
Mid A, decided by:
- Shape:
bin/rule_hook/rules.py:103heavy: about 50 lines of command-word parsing sit in the module that holds the rule table, while rules of similar weight have their own modules; line 126 picks docker's index by a nested conditional
Inherited architect, decided by:
- Shape:
bin/rule_hook/rules.py:103heavy: argv parsing for docker and mutation tools joins the rule registry, breaking the pattern of a sibling module per non-trivial rule, and rules.py now mixes registry, file checks and a shell parser - Shape:
bin/rule_hook/rules.py:126heavy: a conditional nested in another decides where docker sits in the words, which a 3am reader unpicks by hand
2026-10-23T12:49:21Z, PR #370 at 686c1dc, against 64d4386
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Locality:
bin/rule_hook/heavy_command.py:1heavy_command: the file name and docstring lead a cold reader to the code that decides the run-heavy rule, which sat in the middle ofrules.py:103-150before - Shape:
bin/rule_hook/rules.py:8rules: the file drops from 256 to 207 lines and reaches the heavy-command logic through one import - Shape, Locality:
bin/rule_hook/heavy_command.py:28heavy: docker's position is two named steps,reachableanddocker_at, where a nested conditional stood atrules.py:126
Inherited architect, decided by:
- Navigation, Shape:
bin/rule_hook/heavy_command.py:1heavy_command: a file name in therule_hook/listing points straight at the unit, which the base buried in the 256-linerules.py:103-150 - Locality:
bin/rule_hook/heavy_command.py:27heavy: the rule "past a launcher docker can come anywhere, otherwise only first" reads from two plain lines, where the base held a nested conditional - Locality, Shape:
bin/rule_hook/heavy_command.py:5LAUNCHERS: the tables sit beside their only consumer, so extending them needs no search for other readers
2026-10-23T23:00:27Z, PR #372 at 899e8ed, against d1733a2
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | better | same | better |
| Inherited architect | worse | better | same | same | same |
| Maintainability senior | same | better | better | same | better |
| Junior A | same | better | better | same | better |
Mid A, decided by:
- Locality, Shape:
bin/shell_command.py:102launched: one place finds the tool a command runs past its launchers, where heavy held launched, reachable, docker_at and a MUTATORS special case at once - Shape, Locality:
bin/rule_hook/heavy_command.py:12heavy: two cases and one early exit, with run-heavy read as a launcher instead of a stray word - Shape, Navigation:
bin/docker_command.py:1docker_command: named by what it holds and serving both docker rules, but at bin's root while only rule_hook reads it, against README's map - Locality:
bin/docker_command.py:54helped: unlimited leans on helpers defined below it, and helped finds the subcommand by its first occurrence
Inherited architect, decided by:
- Locality:
bin/rule_hook/heavy_command.py:12heavy: four lines on launched and docker_subcommand, though LAUNCHERS now changes both docker rules together - Navigation:
bin/docker_command.py:1docker_command: README says root modules serve more than one tool, yet it serves only rule_hook, so a reader looks in rule_hook/ first - Locality:
bin/docker_command.py:32unlimited: the docstring states that the container's own arguments are read, and why python -m pytest is no limit - Shape:
bin/shell_command.py:1shell_command: the module now does two things, parsing and launcher reading, which only rule_hook uses
Maintainability senior, decided by:
- Locality, Shape:
bin/shell_command.py:102launched: one function, returning a Launch, finds the tool past its launchers, where heavy did it inline with index arithmetic - Locality, Shape:
bin/rule_hook/heavy_command.py:12heavy: one branch, docker or a mutation tool, with the module no longer holding general shell and docker parsing - Shape:
bin/docker_command.py:1docker_command: README says root modules serve more than one tool, yet only rule_hook imports it - Shape:
bin/docker_command.py:32unlimited: the entry point sits between helpers, and unlimited and unlimited_container differ by an unstated input - Locality:
bin/docker_command.py:54helped: reads --help by position while heavy's LIGHT_FLAGS reads it anywhere, and neither says whether that is meant
Junior A, decided by:
- Locality, Shape:
bin/shell_command.py:102launched: the steps that find a command's tool past its launchers live in one function both rules share - Shape, Locality:
bin/rule_hook/heavy_command.py:12heavy: from 13 lines mixing three parsings to 8 with a clear docker-or-mutator branch - Shape:
bin/docker_command.py:1docker_command: a truthful name and docstring, though it sits at bin's root rather than beside rule_hook's rules - Locality:
bin/rule_hook/rules.py:140command_breaches: the new rule plugs in beside run-heavy as one parallel if
2026-10-24T09:53:49Z, PR #373 at 3ac8c79, against eb7b506
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | better | same |
| Inherited architect | same | worse | worse | same | worse |
| Maintainability senior | same | worse | worse | same | worse |
| Junior A | same | worse | same | same | same |
Mid A, decided by:
- Locality:
bin/rule_hook/rules.py:154unlimited_container: rule-hook now encodes how run-heavy's docker puts itself on PATH, joined to it by a one-line comment - Locality:
bin/run_heavy/main.py:65capping: a stand-in docker on PATH and RUN_HEAVY_MEMORY make a container's limits depend on the environment, though the help spells it out - Self-sufficiency:
bin/docker_command.py:24size: named predicates replace a negative-lookahead regex, and swap_limit says what -1 and 0 mean - Self-sufficiency:
bin/docker_command.py:52subcommand_at: positions replace a first-match lookup that took an option's value for the subcommand - Shape:
bin/docker_command.py:84capped: the module grows from 5 functions to 13, and one line packs four conditional limit lists
Inherited architect, decided by:
- Locality, Shape:
bin/rule_hook/rules.py:154unlimited_container: index arithmetic finds the words before docker, and the rule encodes how another package's PATH shim works - Locality:
bin/rule_hook/rules.py:165resets_path: reads as a guess about launchers rather than a model of them - Locality:
bin/run_heavy/main.py:65capping: whether a docker run gets limits depends on PATH order and a variable an ancestor set - Shape:
bin/docker_command.py:39unset_swap: named unset though -1 is unlimited, beside a generic zero, in a module that now reads and rewrites commands - Shape:
bin/docker_command.py:24size: replaces a hard negative-lookahead regex, offsetting in part
Maintainability senior, decided by:
- Locality:
bin/rule_hook/rules.py:154unlimited_container: repeats how run-heavy's PATH trick works, so a change to run-heavy silently changes the rule, and the index arithmetic takes a second read - Locality:
bin/machine_load/memory.py:20Memory.slot_share: tuning the worker threshold ROOM_SHARE now also changes every capped container's memory, while its comment names only the worker - Locality:
bin/run_heavy/main.py:65capping: the share reaches the docker shim through an environment variable and a PATH entry, hidden state between two processes - Shape:
bin/docker_command.py:24size: the module grows to 13 functions that read and rewrite commands, with predicates named as nouns - Shape:
bin/docker_command.py:84capped: one line crosses the two limits' values over and needs the docstring
Junior A, decided by:
- Locality:
bin/rule_hook/rules.py:154unlimited_container: rule-hook copies run-heavy's PATH trick, and nothing in run_heavy points back to it - Locality:
bin/run_heavy/main.py:65capping: one process sets RUN_HEAVY_MEMORY and PATH, and another reads them, so a docker run spans two processes
2026-10-28T01:47:11Z, PR #371 at 51faf71, against 3bde9cd
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | same | same | same |
| Inherited architect | same | same | same | same | same |
Mid A, decided by:
- Locality, Overall:
bin/loop_claude/loop/reload.py:28LoopRun.unfinished_run: its name is one word fromRoute.unfinished, and the two reset at different times - Locality:
bin/loop_claude/loop/main.py:214closed_cycle: the order of archiving the step-up before the limit is checked has to be known, and theorfallback contradicts the comment above it - Shape:
bin/loop_claude/cycle/worker.py:15AUTO_COMPACT_WINDOW: the constant named for the window holds the environment variable's name - Self-sufficiency:
bin/loop_claude/loop/main.py:151UNFINISHED_LIMIT: the number rests on archive evidence the reader can't see - Navigation:
bin/loop_claude/cycle/run.py:82archived_stop: a loop-level stop goes through a cycle-level helper
Inherited architect, decided by:
- Locality, Overall:
bin/loop_claude/loop/main.py:214closed_cycle: the order of its steps is easy to break, and theorfallback contradicts the comment above it - Locality:
bin/loop_claude/loop/reload.py:28LoopRun.unfinished_run: a rerun sets it to 0 only through the dataclass default, which the call site atloop_run_fromdoesn't say - Navigation:
bin/loop_claude/cycle/judge.py:150run_judge: the path from a rejected verdict to exit 1 crosses three files, and only some misquotes count - Navigation, Shape:
bin/round_titles.py:8CAP: the fourth-round cap lives in a module named for commit titles - Self-sufficiency:
bin/loop_claude/loop/main.py:1module docstring: the exit list maps a code to a cause, but no bullet links to the unit that raises it
2026-10-28T11:36:42Z, PR #386 at 722dd5e, against 243a9d8
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | same | same | same |
| Inherited architect | same | worse | better | same | same |
Inherited architect, decided by:
- Locality:
bin/routing_figures/measure.py:155costed: one message's dollars now depend on every message from the same model in the chunk and on the worker's last cost-state entry, so pricing one message means holding the whole chunk's split - Locality:
bin/routing_figures/measure.py:125measured: the one-linelacking=count is explained only in theChunkFiguresdocstring and inbasis - Shape:
bin/routing_figures/measure.py:85Message: named fields replace unnamed(at, tools, spent)tuples and six positional arguments - Shape:
bin/routing_figures/prices.py:33listed: list-price arithmetic gets its own small module with a name that fits
2026-10-28T16:23:54Z, PR #387 at 0447f2e, against cde7720
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | worse | same | worse |
| Inherited architect | same | worse | same | same | same |
| Maintainability senior | same | worse | same | same | same |
| Junior A | same | worse | same | same | worse |
Mid A, decided by:
- Locality, Shape, Overall:
bin/rule_hook/main.py:117read_breaches: the repeat-read rule's file and transcript reads sit in main.py, apart from its logic in repeat_read.py, unlike the other rules, so changing the rule needs both files open - Locality:
bin/rule_hook/repeat_read.py:72shown_lines: whether a Read is checked turns on a text comparison of the mtime with transcript stamps, which holds only while main.modified formats the mtime as Claude Code stamps entries - Locality:
bin/rule_hook/repeat_read.py:29repeat_breaches: needs lines read one past the last shown line, a promise main.py's caller keeps with nothing enforcing it - Shape:
bin/rule_hook/repeat_read.py:87results_on: returns unnamed four-part tuples that callers unpack by position - Shape:
bin/rule_hook/repeat_read.py:103written: returns a path, whilewrittenin main.py names the text that may carry an override marker - Locality:
bin/rule_hook/main.py:84breaches_of: a Bash call now also reads the transcript and the file system through read_breaches - Shape:
bin/claude_code/entries.py:100ToolResult: tool-result parsing has one named home, which shrinks measure.py'sresultsto one line
Inherited architect, decided by:
- Locality:
bin/rule_hook/main.py:117read_breaches: the rule's I/O sits in main.py and its logic in repeat_read.py, with two contracts across the seam, the mtime stamp's format and reading one line past the last shown, that only docstrings in the other file state - Locality:
bin/rule_hook/repeat_read.py:72shown_lines:askerswitches on a Read-only cut-off whose reason lies in a comment and the rule text, the order-dependent part of the rule - Shape:
bin/rule_hook/repeat_read.py:87results_on: untyped positional tuples and dense conditionals at :46 and :105, offset by the sharedToolResultparser inbin/claude_code/entries.py:100and measure.py's dropped duplicate
Maintainability senior, decided by:
- Locality:
bin/rule_hook/main.py:117read_breaches: the rule's I/O sits in main.py and its logic in repeat_read.py; the countmax(shown) + 1at :126 sets what an open-endedcatmeans in the other file - Locality:
bin/rule_hook/repeat_read.py:72shown_lines:changed > touchedcompares stamps as text, which holds only whilemodifiedin main.py formats the mtime as the transcript does;changedandaskermade it the slowest unit to read - Locality, Shape:
bin/claude_code/entries.py:107tool_result: one named parser replaces measure.py's privateresult_text, thoughmeasure.py:142agent_callsstill readspart.get("id")itself
Junior A, decided by:
- Overall, Locality:
bin/rule_hook/main.py:117read_breaches: the rule's entry point, stat and file reads sit in main.py while its logic sits in repeat_read.py, so changing the rule needs both files - Locality:
bin/rule_hook/repeat_read.py:29repeat_breaches: needs lines through one past the last shown and a mtime formatted as a transcript stamp, both set up by main.py - Locality, Shape:
bin/routing_figures/measure.py:205results: one line over the sharedtool_result, wherecde7720held a private copy of the parsing - Shape:
bin/rule_hook/repeat_read.py:87results_on: an unnamed four-field tuple unpacked by position, andwrittennames a path here but the override text in main.py
2026-10-28T17:21:57Z, PR #390 at 3ef4790, against 47a4075
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | same | same | better |
| Inherited architect | same | better | same | better | better |
Mid A, decided by:
- Navigation, Locality, Overall:
bin/rule_hook/round_cap.py:8round_cap_breaches: a refusal by round-cap is traced to a module named after the rule, wired liketodo_check.py, where the base mixed it intocommand_breaches - Locality, Overall:
bin/git_repo.py:81branch_titles: round-count and the hook read the branch's round titles through one function, so they agree by construction - Locality:
bin/rule_hook/override_log.py:104print_overrides: the second definition of the default branch,local_default_branch, is gone
Inherited architect, decided by:
- Locality, Self-sufficiency, Overall:
bin/git_repo.py:81branch_titles: the hook and round-count now count rounds one way, with first parents, where the base's two rules disagreed - Locality, Overall:
bin/rule_hook/override_log.py:104print_overrides: one default-branch lookup instead of a quietly different local one - Self-sufficiency, Overall:
bin/rule_hook/round_cap.py:24earlier_titles: the docstring states both fail-open cases, and the refusal names the counted commits and both ways out - Self-sufficiency:
skills/larv-review/SKILL.md:78"Cap": the ask, yes, one-more-round loop is gone, though the bullet now packs four ideas - Locality:
bin/git_repo.py:63default_ref:guarded=Truemeans both safe git options and never asking origin, kept at each call site; where it rots first - Locality:
bin/rule_hook/main.py:113bash_breaches: round-cap sits apart from its sibling round-title-counts incommand_breaches, a split no comment states
2026-10-28T17:54:03Z, PR #391 at f1d749e, against 9014fd0
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | better | better | same | same | better |
Mid A, decided by:
- Locality, Overall:
bin/loop_claude/effects.py:93say: every message goes through one function that adds the time, picks stderr or stdout and always flushes; before, each call site chose its own stream and flush, so the order of stdout and stderr lines depended on hidden buffering - Locality:
bin/claude_code/trust.py:10trust: reports through anotecallable the caller passes, so the call shows where its output goes - Locality:
bin/transcript_archive/write.py:14archive: same, but its defaultnotesilently drops thetranscript: …line for a caller that leaves it out - Self-sufficiency:
bin/loop_claude/effects.py:40Effects.local_time: the new comment gives the format and why it is local time - Self-sufficiency:
bin/loop_claude/loop/main.py:289plan_item: the new comment sends a cold reader to the plan, and "one injected effect" is unclear besidesay - Shape:
bin/loop_claude/cycle/run.py:85archived_stop: the samelambda text: say(text, effects)is written out five times andarchivegrows to six parameters
Inherited architect, decided by:
- Navigation, Locality, Overall:
bin/loop_claude/effects.py:93say: about 30printcalls over nine files, disagreeing on flushing and streams, now go through one function whose docstring states the contract - Locality:
bin/claude_code/trust.py:10trust: wrote tosys.stderritself before; now takes anotecallback, so the caller decides where its words go - Locality:
bin/transcript_archive/write.py:14archive: the same change, but its silent defaultnotedrops thetranscript: <path>line for a new caller without a word - Shape:
bin/loop_claude/cycle/run.py:85archived_stop: the same lambda is copied seven times acrossrun.py,loop/main.pyandcycle/interrupt.py, which roughly cancels the cleaner output path - Self-sufficiency:
bin/loop_claude/loop/main.py:289plan_item: its comment sends the reader to the plan and reads oddly now that every message goes throughsay
2026-10-28T21:50:16Z, PR #392 at 6e12ee9, against 77bc635
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Locality, Shape:
bin/branch_chunks/cut.py:59cut_branch: the olderscore_outcomes/main.pycut the branch with index arithmetic over raw log lines, each chunk a(short, short, int, Landing)tuple read aschunks[-1][2], with three helpers agreeing on hash slices; the newer typedCommit,SpanandBranchdataclasses make one pure cut, the landing rule sits in its ownlanding, and both tools share it - Navigation:
README.md:128the import-root paragraph: the older packed every package-to-package dependency into one long paragraph; the newer gives each package a bullet, sorted by name, so going from a package to what it serves and imports is one scan - Shape, Locality:
bin/branch_chunks/read.py:7branch_chunks: every git call sits in this short shell whilecut.pyandfixes.pystay pure, andfixes.py's docstring says what fixed later means before any code; the two tools only seeBranchChunk.fixed_by - Shape:
bin/round_titles.py:34rounds_past_first: the older computed rounds past the first in two places, one mixed into title extraction; the newer has one helper, andmeasure.py:229commit_titlesdoes only what its name says - Locality, Shape:
bin/loop_claude/effects.py:31Effects.git: usesgit_repo.Gitinstead of a narrower local protocol, so one definition says what a git runner is; a small cost is thatgit_repo.Gitandgit_repo.Runare two neighbouring types for something that runs git - Self-sufficiency:
bin/routing_figures/measure.py:336fixes_per_chunk: its comment says "A mean, where item 247 named the median", which only reads fully withtodo.mdopen; the new docstrings infixes.pyandscore_outcomes/main.py:27explain later fixes in place, and the two balance
Inherited architect, decided by:
- Shape, Locality:
bin/branch_chunks/cut.py:59cut_branch: the olderchunk_logreturned a positional tuple read by index; the newerCommit,SpanandBranchare frozen dataclasses with named fields, andlandingandlabelare small functions each readable and changeable alone - Shape:
bin/score_outcomes/results.py:40results_of: the older took six positional arguments its caller built from rev-range strings such asf"{first}^1..{first}^2"; the newer takes oneBranchChunkand the numstat, andread_resultsis a short loop without rev-range arithmetic - Navigation, Locality:
bin/routing_figures/measure.py:229commit_titles: the older wrote the rounds-past-first sum twice, once mixed with reading the transcript; the newer has one definition inround_titlesused by both tools, and reading titles is its own unit - Navigation:
README.md:128the module list: the older described the module dependencies in one run-on paragraph; the newer gives one bullet per module, includingbranch_chunks/naming who uses it - Shape:
bin/routing_figures/measure.py:257totals: the older builtFiguresfrom eight positional arguments; the newer passes them by keyword, which matters now that it gainedunmatchedandunread - Locality:
bin/git_repo.py:48Git: the protocol moved out ofloop_claude/effects.pyintogit_repo, anddefault_refandbranch_chunkstake the runner as an injectedrun, so the guarded and timed path reads at the call site - Locality:
bin/branch_chunks/fixes.py:50with_fixes:branch.owner[commit] > change.chunkmeans a later chunk fixed an earlier one only because chunks are ordered newest first, which onlyBranch's andchunk_log's docstrings state; not enough to move the ruling - Self-sufficiency:
bin/routing_figures/measure.py:336fixes_per_chunk: explains its choice by pointing to "item 247" in the plan, whilelater_fixes_ofandfixes.py's docstring state their rules in full, so it is mixed
2026-10-29T00:40:36Z, PR #394 at bbb33f5, against b743228
Ruling: worse
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | worse | same | worse |
| Inherited architect | same | worse | worse | same | worse |
Mid A, decided by:
- Locality, Overall:
bin/loop_claude/cycle/interrupt.py:36put_question, withbin/loop_claude/cycle/interrupt.py:90talk: the ask is published input_question, whosefinallyleaves it listed on aTalk, andtalk's ownfinallyremoves it one state later; changing either means following the ask's lifecycle across both and theTalkingstate, where at the base each did one local thing - Locality, Shape:
bin/loop_claude/cycle/interrupt.py:98opened: its call toshownis also the go-aheadclaim_talk(bin/loop_claude/waiting.py:192) polls for, so a function named for showing a question is one step of a handshake between two processes; only the comment above it andwaiting.py's docstring warn - Shape:
bin/loop_claude/loop/routing_check.py:7raise_calls: the imports now reach intocycle.interruptforclaude_args,shownandwithdrawnplus thewaitingtypes, and it builds anAskwhose session is its own id - Shape:
bin/loop_claude/effects.py:12the import ofAnswer: the low-level effects module, andstates.py, now depend on the domain modulewaiting.pyfor their types - Shape, Self-sufficiency:
bin/loop_claude/cycle/states.py:12TypedAnswer: its docstring still says typed at loop-claude's terminal, though an answer from loop-inbox becomes one too, andTalkinglost its docstring; minor, outweighed bywaiting.py's docstring - Shape:
bin/loop_claude/cycle/interrupt.py:22WorkerSession: bundles the base's seven loosetalkarguments and sixhand_overarguments, which reads better and keeps Shape only mildly worse - Navigation:
README.md:102the loop-inbox entry: the new tool is signposted from the README,loop-claude --helpandloop-inbox --help, so a symptom crosses more files but each is named where you would look
Inherited architect, decided by:
- Locality:
bin/loop_claude/cycle/interrupt.py:36put_question, withbin/loop_claude/cycle/interrupt.py:90talk: on aTalkthe ask's file stays up while the answer passes throughrun.ask,TalkChosen,transitionandTalking, andtalk'sfinallytakes it down; to be sure it is always removed, a reader holds the state machine across three files - Locality:
bin/loop_claude/waiting.py:192claim_talk: loop-inbox waits 5 s forOpenInInbox, which the loop shows only after five units in three other files, so step 2 of the protocol is spread out and the timeout rests on that path staying quick - Locality:
bin/loop_claude/cycle/interrupt.py:78shown: confirming its docstring's failure path, loop-inbox gives up and the loop stops, takes four steps across two processes and two files - Shape:
bin/loop_inbox/main.py:39the import ofhand_over:interrupt.pyundercycle/now also holdshand_over,claude_args,shownandwithdrawn, whichrouting_checkand loop-inbox import, so its name and folder no longer say who depends on it, whilewaiting.pyholds only the rawpublishandwithdraw - Shape:
bin/loop_claude/cycle/states.py:12TypedAnswer: its docstring still says typed at loop-claude's terminal;Talkinglost its docstring, andstates.pyandeffects.py:12now import from the inbox's file format - Locality:
bin/loop_claude/cycle/run.py:102step, theTalkingbranch:WorkerSessiongathers the base's seven loose arguments andclaude_commandbuilds the resume command in one place, a smaller gain outweighed by the above - Self-sufficiency:
bin/loop_claude/waiting.py:1the module docstring: documents the file protocol in three numbered steps, so nothing new needs a document beside it; the cost is how far apart the pieces are
2026-10-29T00:46:50Z, PR #394 at bbb33f5
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6.5 | 6 | 6 |
| Mid A | 7 | 5.5 | 6.5 | 6 | 6 |
| Maintainability senior | 6 | 5 | 6 | 5.5 | 5.5 |
| Inherited architect | 7 | 6 | 5.5 | 7.5 | 6.5 |
| Mean | 6.75 | 5.50 | 6.13 | 6.25 | 6.00 |
Junior A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: four interacting rules, with helpers in ladder.py, route.py and escalation_mark.py, so predicting one rung means holding the rung, the unfinished count, the mark, the cap and taking_in at oncebin/loop_claude/cycle/prompts.py:55review_launches: English instructions assembled from routing state and review_choice.chosen, whose order matters and whose result only the docstring examples showbin/loop_claude/loop/reload.py:45decoded_as: a generic recursive decoder with REJECTED and RETIRED sentinels and an unstatedfitsconvention, where a small mistake breaks the reload's hand-over of statebin/loop_claude/cycle/judge.py:150run_judge: a misquote ends three different ways, with a verdict precedence only the comments inGroundsresolvebin/rank_todo.py:410blockers_first: recursive blocker placement with loop detection, and every scoring function branches on the old and new table shapesbin/chunk_pick.py:11LastCut: fields typed withFailureput errors inside the data, and the renamed-release rule lives only in named_releasesskills/larv-review/SKILL.md:72"Round cap": dense rules held together by cross-references, so one question took four filesbin/rule_hook/rules.py:160unlimited_container: a PATH heuristic about run-heavy decides whether a container is capped
Mid A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: one rule raises, steps once, caps, moves the escalation mark and takes in E + I, readable only with routing.py, escalation_mark.py, route.py and ladder.py open togetherbin/loop_claude/cycle/run.py:167ask: which rung a step-up is charged to travels as a StepUp inside the states, action at a distance the comments at 145 and 170 resolve twicebin/loop_claude/loop/reload.py:45decoded_as: a recursive reflective decoder with sentinels and a duck-typedfits, whose consequence is spelled out only on LoopRunbin/rank_todo.py:428place: recursive blocker placement, with the old and new table shapes running through read_scoring, weigh, score and cited_goals_failurebin/comprehension_history.py:207labelled_bullets: a stateful line parser whose branches serve formats from before PR #232 and #255bin/loop_claude/cycle/judge.py:164quotes_checked: a misquoted reply ends three ways, plus grounded's ellipsis matching, found only by tracing each callerbin/model_routing/route.py:84route_from: reads three spellings of one value, and nothing says when the legacy form can gobin/review_replay/main.py:234redirected: rewrites archived prompt paths where order matters, and guesses a round's commit from author times
Maintainability senior, hardest first:
bin/loop_claude/routing.py:76after_unfinished: five rung concepts, each defined elsewhere, plus the unfinished counter, a mark counted in transcript entries and taking_in, replayed by hand on examplesbin/loop_claude/loop/reload.py:45decoded_as: generic reflection machinery inside a domain package; nothing says when a field becomes RETIRED instead of REJECTEDbin/rank_todo.py:410blockers_first: two table shapes side by side, recursive placement, and weight rules stated together only in the module docstringbin/chunk_pick.py:166named_releases: the next release derived from every todo.md since the tag minus renames, clearer in chunk-preflight's help than at its homebin/comprehension_history.py:130newest_scored: prefix-matched commits, regex headings whose differences matter, and legacy carve-outs citing PR numbers instead of the old shapesbin/loop_claude/cycle/run.py:190archive_stepped: the route update runs in a channel beside the states, so what changes the route has two answers, here and in loop/main.py's after_verdictbin/loop_claude/cycle/prompts.py:55review_launches: which sentences reach a worker's prompt depends on a dozen small functions, and no whole prompt is visiblebin/rule_hook/rules.py:312unlimited_container: slice arithmetic on words, and resets_path counts -i and -u whatever launcher they follow
Inherited architect, would restructure first:
bin/round_titles.py:8CAP: the round cap lives in a module named for commit titles, and nothing says a third reviewer launch is left unguarded on purposebin/loop_claude/routing.py:76after_unfinished: the step-up policy is split from model_routing/ and spread over about six files, so why a worker is on rung X takes two packagesbin/loop_claude/loop/main.py:169run_loop: about twelve steps per cycle and an import fan-in of about 20 modules from 8 packagesbin/loop_claude/waiting.py:76Typed: one name over three concepts, with transcript.py's Typed and states.py's TypedAnswerbin/rung_replay/search.py:1the imports: replay tools reach into loop_claude.cycle.retry and loop_claude.effects, so loop_claude is also the shared headless-claude librarybin/loop_claude/cycle/judge.py:64Reply: a new verdict field touches Reply, verdicts.py, states.py, routing.py, Route and the outcomes archiveREADME.md:129the module map: lists only cross-package imports, leaving the 18 root modules of bin/ without an index
2026-10-29T00:51:59Z, PR #394 rerun at bbb33f5
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6.5 | 6 | 6.5 |
| Mid A | 7 | 5.5 | 6 | 6 | 6 |
| Maintainability senior | 6.5 | 5.5 | 6.5 | 6 | 6 |
| Inherited architect | 7 | 6.5 | 6 | 7.5 | 6.5 |
| Mean | 6.88 | 5.88 | 6.25 | 6.38 | 6.25 |
Junior A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: four rules meet in one expression, so one call means holding item_rung, ESCALATION_CAP, UNFINISHED_PER_STEP and the mark's entries_read while jumping to ladder.py and escalation_mark.pybin/comprehension_history.py:97entries: the module parses its own markdown output, old formats are keyed only to PR numbers, and scored against scoring_runs differ in a way the code doesn't namebin/loop_claude/cycle/run.py:167ask: which route each archived line names before and after a step-up, with the cycle replaced inside run_cycle; the comments at 61, 145 and 170 resolve itskills/larv-review/SKILL.md:31"Passes": dense rules pointing out to CLAUDE.md, rules/, the standards and --help, with the re-entry table and the round cap held togetherbin/rule_hook/round_cap.py:24earlier_titles: returns an empty list when git fails, so a git failure turns the round cap off without saying sobin/rank_todo.py:568score: both table shapes run through the whole 659-line file, and only rules/todo-items.md says what C, A and G meanbin/loop_claude/loop/reload.py:45decoded_as: recursive type decoding with sentinels and a duck-typed fits hook, so a hot reload depends on dataclasses in other filesbin/routing_figures/measure.py:174costed: recorded costs scaled across list prices and messages matched to a rung by timestamp, resolved by long docstrings
Mid A, hardest first:
bin/loop_claude/routing.py:263after_unfinished: one condition needs rung, item rung, kept rung, L taken in, the escalation mark and the cap held at once, with rules/todo-items.md and the ladder openbin/rank_todo.py:332weigh: a 658-line file carries a legacy scoring model through almost every function, so each one asks which shape it is inbin/loop_claude/escalation_mark.py:183checked: part of the contract is a sentence spliced into the judge's schema text, so changing English changes behaviourbin/loop_claude/cycle/judge.py:316verdict_from: the verdict precedence is encoded twice, as the order of the ifs and in the QUOTE_RULES prosebin/loop_claude/loop/main.py:169run_loop: builds Cycle in one line of about 330 characters under 25 unordered imports, and plan_item defers to a todo.md item numberbin/routing_figures/measure.py:59ChunkFigures: fields documented in one dense paragraph instead of beside each field, and render builds tables inline in one huge literalskills/larv-review/SKILL.md:72"Round cap": what counts as a round, the auto-reviewer exception and relaunches staying in their round took rereads beside round_titles.py and round_cap.py
Maintainability senior, hardest first:
bin/loop_claude/routing.py:76after_unfinished: one verdict raises the rung, moves the mark, resets or bumps the count and takes in E + I, with invariants in route.py, ladder.py and escalation_mark.py held open at oncebin/loop_claude/cycle/run.py:167ask: the StepUp travels through the Working and Interrupting states while run_cycle swaps Cycle.route out of sight of these handlersbin/rank_todo.py:568score: the old table shape forks about eight functions, and the module docstring describes only the new shapebin/loop_claude/cycle/prompts.py:55review_launches: the worker's prompt is built from conditional fragments, so the text the worker reads appears only in docstring examplesbin/loop_claude/loop/reload.py:45decoded_as: a generic recursive decoder with first-fit union matching, sentinels and a duck-typed fits, which every LoopRun field quietly depends onbin/shell_command.py:44segments: a hand-made state machine over shlex tokens tracks redirect targets, and a change's safety is unclear without the testsbin/chunk_pick.py:109next_release: LastCut fields typed str, None or Failure, and recovering renamed releases from todo.md's history is explained only in passingbin/loop_claude/waiting.py:192claim_talk: a handshake between processes with a timeout and a take-back, resolved by the module's three-step docstring
Inherited architect, would restructure first:
README.md:129the module map: five domain modules sit at bin/'s root because two tools share them, so larv-review's rules spread over larv_review/, round_titles.py, review_passes.py, review_folder.py and comprehension_history.pybin/git_command.py:17commits: parses git commit out of shell words and doesn't run git, beside git_repo.py's git and rule_hook/git.py's guarded gitbin/rung_replay/search.py:1the imports: replay tools reach into loop_claude.cycle.retry and loop_claude.effects, so loop_claude is also a shared librarybin/rule_hook/main.py:79breaches_of: rules.py is both the RULES registry and the checker of about half the rules, while the other half have their own modulesbin/loop_claude/routing.py:76after_unfinished: the step-up policy is spread over about six files across loop_claude and model_routing, where the next routing rule landsbin/round_titles.py:8CAP: enforced only on round commit titles, so a third reviewer launch is guarded by SKILL.md prose alonebin/loop_claude/project/records.py:1RECORD_FILES: "records" means the judge's records here and a tool-kept document in AGENTS.mdbin/loop_claude/loop/main.py:1the module docstring: a 122-line contract restating README and SKILL.md, where drift will come from
2026-10-29T12:15:18Z, PR #394 at 9ebd06e, against b743228
Ruling: worse
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | better | same | worse |
| Inherited architect | same | worse | same | same | worse |
Mid A, decided by:
- Locality:
bin/loop_claude/cycle/interrupt.py:39interrupt: where the base'sput_questionread one terminal line, it waits on the terminal and on an answer file another process writes, and hands a lambda wrappingclaude_opentohuman_typed, so a change means holding the two-process protocol; its docstring is one long sentence over every branch - Locality, Shape:
bin/loop_claude/waiting.py:202claim_talk, withbin/loop_claude/waiting.py:221make_way: /talk from loop-inbox is a three-step handshake across two processes, with a take-back on timeout; order dependence with no counterpart at the base, though the module docstring explains it - Shape:
bin/loop_claude/cycle/interrupt.py:74first_answer:OpenOnTerminal, an ask's state in loop-inbox's file, doubles as the return value for /talk and, at line 48, for a release ready to cut, so the name tells the truth about the file but not the control flow - Shape:
bin/loop_claude/cycle/states.py:34Interrupting: the state machine losesTalkingandTalkChosen, sotransitionhas one state and one event fewer - Shape:
bin/loop_claude/hand_over.py:24hand_over: opening claude on a terminal leavescycle/interrupt.pyfor its own module, ending routing_check's sideways reach into the cycle, andClaudeCommandreplaces six loose arguments - Shape:
bin/loop_inbox/inbox.py:59pressed: the new tool splits into a pure core of screens and keys behind a curses shell, with names that say what they are
Inherited architect, decided by:
- Locality, Overall:
bin/loop_claude/waiting.py:202claim_talk, withbin/loop_claude/waiting.py:221make_way,bin/loop_claude/cycle/interrupt.py:77claude_openandbin/loop_inbox/main.py:146opened: /talk, one call in one process at the base, is a three-step handshake between two processes through a JSON file, over three files in two packages, resting onMAKE_WAY_SECONDS, a take-back after a timeout and pid liveness; the module docstring lays the steps out well - Locality:
bin/loop_claude/effects.py:105alive: whether an ask still counts, and whether the loop still waits on loop-inbox's claude, rests on a pid being alive, and nothing addresses pid reuse, a risk neither caller shows - Locality:
bin/loop_claude/cycle/interrupt.py:39interrupt, withbin/loop_claude/cycle/interrupt.py:94human_typed: the base read straight down,put_questionreturning and the state machine stepping toTalking; now the wait sits insidelisted,first_answerpolls two sources, and the claude wait passes as a lambda; fewer states andhand_over's move out ofcycle/are real gains that don't offset it
2026-10-29T13:05:34Z, PR #394 at 9ebd06e
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6 | 6 |
| Mid A | 7 | 5.5 | 6.5 | 5.5 | 6 |
| Maintainability senior | 7.5 | 6 | 6.5 | 7 | 6.5 |
| Inherited architect | 7 | 6 | 6 | 7 | 6.5 |
| Mean | 7.13 | 5.75 | 6.25 | 6.38 | 6.25 |
Junior A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: five interacting rules in one function, each resting on terms defined in ladder.py, route.py, escalation_mark.py and the charter; the module docstring's outline says which branch, not why the orderbin/loop_claude/cycle/run.py:165ask: the route before and after a step-up travel together, and seeing which one each archive line names meant tracing step_up through states.py; the comments at 61, 143 and 168 resolve itbin/loop_claude/waiting.py:202claim_talk, withbin/loop_claude/waiting.py:221make_wayandbin/loop_claude/cycle/interrupt.py:77claude_open: a three-step handshake between two processes through files, polled against a deadline with liveness checks; the module docstring gives the order, but who writes which file needs both sidesbin/loop_claude/cycle/interrupt.py:39interrupt: a one-sentence docstring over about six outcomes, with the body spread over first_answer, claude_open and human_typed; accurate but too dense for one readbin/rank_todo.py:218rank: about eight failure checks in sequence, and several functions branch on SHAPE versus OLD_SHAPE, a dual path explained only in todo_table.pybin/loop_claude/cycle/judge.py:164quotes_checked: quietly turns a misquoted answer into an Interrupt, and verdict_from's branch order decides the meaningskills/larv-review/SKILL.md:72"Round cap": a dense rulebook of passes, caps, re-entry and severities, part implemented in bin/larv_review and part followed by the agent; stayed too much to hold at oncebin/model_routing/route.py:84route_from: an archived line is JSON or an older regex format, and renaming a field breaks archived lines, guarded only by a comment
Mid A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: kept_rung, item_rung, raised, a one-rung step and a cap check held at once, plus two step triggers and a moving escalation mark; the docstring's index and escalation_mark.py's four steps resolve it partly, after three rereadsbin/loop_claude/cycle/run.py:54run_cycle: a step-up rides on state fields while run_cycle swaps Cycle.route on Ruled, HumanTalked drops it, and ran_on and archive_stepped decide which route a line names; the comments at 61, 143 and 168 explain it, read beside states.pybin/loop_claude/loop/reload.py:45decoded_as: a generic recursive decoder with REJECTED and RETIRED sentinels, duck-typed to Route throughfits, which nothing at Route mentionsbin/rank_todo.py:332weigh: OLD_SHAPE/SHAPE branches spread over weigh, score, cited_goals_failure and goal_weights instead of converted once at read time; why both exist is in todo_table.pybin/loop_claude/cycle/interrupt.py:39interrupt: a docstring of one sentence chain with four where/else turns, then a split across first_answer, claude_open, human_typed and the /talk handshake; waiting.py's module docstring resolves itbin/loop_claude/cycle/judge.py:106verdict_from: an implicit priority order and silent coercions, a misquote becoming an Interrupt and an off-scale L becoming 0bin/rank_todo.py:428place: recursive blocker placement with loop detection, fiddly by natureskills/larv-review/SKILL.md:31"3. Passes": dense rules pointing to CLAUDE.md, review-findings.md, standards/ and larv-review --help, not fully followed
Maintainability senior, hardest first:
bin/loop_claude/routing.py:76after_unfinished: one unit raises the rung to an item's L, steps on every second unfinished verdict or on an escalation below the cap, moves the mark and records E + I, with rules from ladder.py, route.py and escalation_mark.py; the docstring states the rule but each branch needed tracing by handbin/loop_claude/cycle/judge.py:96Grounds: quotes from three sources, each misquote punished differently (Rejected, escalation dropped, answer turned to Interrupt), with the escalation half in escalation_mark.checked; each case commented, but the asymmetry is easy to breakbin/chunk_pick.py:109next_release: the outcomes come from combining is_earliest with landed, and LastCut's X | Failure fields push failure handling into the data; subtle by nature, clear on a second readbin/loop_claude/cycle/run.py:54run_cycle: the route changes on Cycle at a Ruled step-up while step_up travels through the states, and ran_on versus cycle.route decides which rung each archive call charges; the comments at 61, 143 and 168 explain it, but the invariant spans two modulesbin/rank_todo.py:428place: recursive placement with loop detection, with OLD_SHAPE threading through older_columns, parse_weights, goal_term and cited_goals_failure; whether the old shape is still live stayed unclearbin/model_routing/route.py:84route_from: layered backward compatibility (a JSON or legacy Worker line, reload's REJECTED and RETIRED sentinels and afitsfound by name), so a change to Route's fields hits three decodersbin/rule_hook/repeat_read.py:72shown_lines: timestamps compared as text, tool calls paired with results by id and truncation markers handled, closely tied to the transcript format; the comments resolve itskills/larv-review/SKILL.md:31"3. Passes": dense cross-referenced rules, waves, reruns and the re-entry table not holdable in one read
Inherited architect, would restructure first:
bin/rule_hook/round_cap.py:8round_cap_breaches: the round cap is split over round_titles.py, rule_hook/round_cap.py, larv_review/main.py and SKILL.md, and guards onlyAnswer <pass> roundcommit titles, so a third reviewer run slips through on a launch without a round commit, a misspelled pass name or a slow git; nothing in the code says the cap is advisory at launch timebin/loop_claude/cycle/judge.py:186grounded: fuzzy quote matching against a transcript rendered in escalation_mark.py and normalised in claude_code/transcript.py, three modules that must stay in step; where it rots firstbin/round_titles.py:8CAP: thebin/root holds about 55 peers, with review mechanics, shell parsing, plan reading and plumbing side by side, the README grouping only part; the worst levelbin/git_command.py:17commits: the name says it runs git, but it pulls commit messages from parsed shell words, beside git_repo.py, which runs git, and rule_hook/git.py, a third wrapper; three opens to sort outbin/loop_claude/routing.py:76after_unfinished: routing reaches over routing.py, model_routing, cycle/prompts.py and, through environment variables, rule_hook/reviewer_rung.py; hard because the policy is, and documentedbin/loop_claude/loop/main.py:1module docstring: 120 lines of contract that restate CLAUDE.md's Vocabulary, AGENTS titles and the README's Tools, so a vocabulary change must land in four placesskills/larv-review/SKILL.md:31"3. Passes": about twelve interacting concepts under one heading, the prose twin of the bin/ rootbin/loop_claude/cycle/states.py:88transition: "ruling" names both the judge's ruling and the comprehension paired ruling, "project" a rung, a package and project_key, "chunk" three tools
2026-10-29T13:05:37Z, PR #394 rerun at 9ebd06e
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6 | 6 |
| Mid A | 7 | 6 | 6 | 5.5 | 6 |
| Maintainability senior | 7 | 6 | 7 | 6 | 6.5 |
| Inherited architect | 7 | 6.5 | 5.5 | 6.5 | 6 |
| Mean | 7.00 | 6.00 | 6.13 | 6.00 | 6.13 |
Junior A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: the unfinished count, a step every two verdicts, a two-rung cap, the escalation mark and the item's rung interact, unpredictable on one read; the docstring and rules/todo-items.md gave the words, the arithmetic only made sense after tracing loop/main.pybin/loop_claude/escalation_mark.py:53checked: the mark is an entry count that becomes a[step]line in the judge's prompt, which the grounding check in judge.py then relies on; one concept over three files, neither naming the otherbin/rank_todo.py:428place: recursive placement with loop detection, and the OLD_SHAPE/SHAPE branches through weigh, goal_term, score, parse_weights and cited_goals_failure; the spec docstring explains why, the recursion stayed hardbin/loop_claude/loop/reload.py:45decoded_as: a recursive type-directed decoder with REJECTED and RETIRED sentinels and a duck-typedfits, generic machinery whose reason lives at exec_if_editedbin/comprehension_history.py:150scoring_runs: a near-twin ofscoredat 159 that a reader would mix up, with branches for formats from before PR #232 and #255 that need history the reader doesn't havebin/model_routing/review_choice.py:21chosen: many near-synonym ladder helpers and an opaque "higher" loop; the docstrings explain edge cases, not the shapebin/rule_hook/repeat_read.py:72shown_lines: timestamps compared as strings and truncation heuristics, correct only under assumptions not stated therebin/loop_claude/loop/main.py:1module docstring: about 120 lines of contract before any code, accurate but costly, citing a plan item the reader may not open
Mid A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: the kept rung, the item's rung, the cap, the mark's step and whether E + I was taken in, held at once, with meanings spread over routing.py, ladder.py, route.py and escalation_mark.py; no single place states the rule as a wholebin/loop_claude/loop/reload.py:45decoded_as: a generic recursive decoder signalling through REJECTED and RETIRED and duck-typingfits; why it exists is stated, which types reachfitstook a grepbin/rank_todo.py:332weigh: both table shapes branch inside weigh, score and cited_goals_failure, so a scoring change must be made right in several places; nothing collects the shape logicbin/routing_figures/measure.py:104measured: many collaborators and a proportional cost scaling, a ChunkFigures docstring defining every field in one paragraph, and rung_at comparing timestamps as strings without saying why that holdsbin/loop_claude/cycle/prompts.py:55review_launches: sentences that branch on route, project rung and taken-in state, a second copy of routing.py's rung rules in English; the worked examples helpbin/model_routing/route.py:84route_from: JSON and legacy regex lines both parse, RETIRED covers removed rungs, and nothing says when a form may be droppedbin/loop_claude/loop/main.py:1module docstring: 120 lines of contract before the first import, which routed and closed_cycle only make sense againstbin/rule_hook/repeat_read.py:72shown_lines: rebuilds which lines a Read showed from transcript results with truncation heuristics and string-compared timestamps; the module docstring mostly resolves it
Maintainability senior, hardest first:
bin/rank_todo.py:218rank: the largest module, with every scoring step branching on OLD_SHAPE versus SHAPE and checks reaching into README.md and docs/decisions.md; the format docstring doesn't say why two shapes coexistbin/loop_claude/routing.py:76after_unfinished: three rules in one function, followed only with route.py, escalation_mark.py, verdicts.py and ladder.py open; the docstrings and escalation_mark's overview resolve it eventuallybin/loop_claude/loop/reload.py:45decoded_as: callskind.fitsby naming convention, so Route.fits is action at a distance and a rename breaks the reload silently; the LoopRun comment warns about fields, not this hookbin/chunk_pick.py:109next_release: the earliest release after the tag depends on renames across historical todo.md texts; chunk_preflight restates the rule, the rename tracking stays subtlebin/loop_claude/cycle/judge.py:150run_judge: three kinds of ungrounded evidence end three ways, Rejected and retried, an Interrupt, or a silently dropped escalation; the Grounds comments resolve itbin/model_routing/review_choice.py:21chosen: thehigherloop records the lowest E + I per rung, and itsorclause needs replay_applies and reviewers_rung from ladder.py; one terse commentbin/loop_claude/cycle/prompts.py:104escalation: sentences assembled from ReviewChoice over three cases, the output described only by docstring examplesbin/rank_todo.py:428place: recursive topological placement with path-based loop detection, traced by hand
Inherited architect, would restructure first:
bin/rule_hook/round_cap.py:24earlier_titles: the round cap holds only at commit time, nothing refuses a third launch, and when git fails or no default branch is recorded it fails open; three files to establish thatbin/round_titles.py:8CAP: thebin/root holds about 60 entries, launchers, tool packages, shared modules and single-file tools at one level, told apart only by naming; the worst levelbin/loop_claude/effects.py:36Effects: 18 callables in loop-claude's home, yet imported by review_replay, loop_inbox and routing_check, so the README's claim that shared code lives at the bin/ root doesn't hold, and every feature adds a field every sharer pays forbin/git_command.py:17commits: reads as "run a git command" but parsesgit commitout of shell words, one of three git modulesbin/rule_hook/rules.py:139command_breaches: some checks sit here, others in their own modules joined by main.py's bash_breaches, so adding a rule needs a choice the code doesn't explainbin/rule_hook/reviewer_report.py:10REVIEWERS: a second source of truth for who is a reviewer, beside review_passes.py's table of the same namebin/loop_claude/routing.py:76after_unfinished: "routing" spans loop_claude/routing.py, model_routing, routing_check, routing_figures and rung_replaybin/loop_claude/loop/main.py:169run_loop: a fan-in of about 25 modules
2026-10-29T19:59:03Z, PR #397 at 1723469, against 77ab11e
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | better | same | same |
| Inherited architect | same | worse | better | same | same |
Mid A, decided by:
- Locality:
bin/loop_claude/loop/main.py:182run_loop: the line aftershown_runningrebindspaths.waitingtolisted.folder, a private unlisted folder when the lock can't be taken;interruptandrouting_checkreadpaths.waitingwith no way to see the swap - Locality:
bin/loop_claude/waiting.py:123listed: works only when the caller passes the folderrunning_loops.shown_runningyields; in the base the ask carried its own liveness check - Locality:
bin/loop_claude/cycle/prompts.py:72review_launches: types out the "Markdown no agent follows" list by hand instead of deriving it fromINSTRUCTION_FILESandINSTRUCTION_FOLDERSinbin/rule_hook/reviewer_rung.py:22 - Shape:
bin/loop_claude/verdicts.py:6TakenIn: three loose fields on each verdict become one named value, andtaken_in_frominbin/loop_claude/cycle/judge.py:139takes the clamping out ofverdict_from - Navigation, Self-sufficiency:
bin/loop_claude/running_loops.py:1module docstring: leads from "loop-inbox doesn't show my loop" to the mechanism as directly as the base's pointer towaiting.py;session_lock.pystill sends the reader todocs/decisions.mdfor why lock files stay
Inherited architect, decided by:
- Locality:
bin/loop_claude/loop/main.py:182run_loop: rebindspaths.waitingtolisted.folder, a throwawayunlisted/<id>folder where another loop holds the lock;cycle/interrupt.pyandloop/routing_check.pywrite questions to a folder they can't see was swapped, explained only atListed.folder - Locality:
bin/loop_claude/running_loops.py:3module docstring: liveness moved from oneAsk.pidfield to a flock shared bysession_lock.py,running_loops.pyandwaiting.py, andlistedhas a precondition stated only in its docstring - Locality:
bin/loop_claude/cycle/prompts.py:72review_launches: its "Markdown only" list repeatsINSTRUCTION_FILESandINSTRUCTION_FOLDERSinbin/rule_hook/reviewer_rung.py:22and the skill's "Markdown only" section, with nothing tying them together - Locality:
bin/loop_claude/running_loops.py:83locked_soon: the flock lapses for a moment when the reload execs into the edited script inside thewithatbin/loop_claude/loop/main.py:185, and no comment says so; the nestedExitStacktakes more than one read - Shape:
bin/loop_claude/verdicts.py:6TakenIn: one value replaces thehighest_*_taken_inpair repeated across three verdicts, andtaken_in_fromchecks the ranges in one place - Shape:
bin/rule_hook/reviewer_rung.py:22prose: the name hides "Markdown no agent follows", andmarkdown_onlyruns git inside a hook module that was pure - Navigation:
bin/loop_claude/session_lock.py:1module docstring: small, well-named modules lead to the protocol quickly, against liveness now spread over three files - Self-sufficiency:
bin/loop_claude/session_lock.py:1module docstring: a new pointer todocs/decisions.md, offset by the in-place explanations atListed.folderandfetched_todo
2026-10-29T20:01:40Z, PR #397 at 1723469
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6 | 6 |
| Mid A | 7 | 6 | 6 | 6 | 6 |
| Maintainability senior | 7 | 6 | 6 | 6 | 6 |
| Inherited architect | 6.5 | 6 | 5.5 | 7 | 6 |
| Mean | 6.88 | 5.88 | 5.88 | 6.25 | 6.00 |
Junior A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: eight names held at once to follow one call; the docstring resolves it only afterescalation_mark.pyandrules/todo-items.mdbin/rank_todo.py:568score: every scoring step branches on SHAPE against OLD_SHAPE, so each function reads twice; nothing says when OLD_SHAPE can gobin/loop_claude/cycle/run.py:56run_cycle: the route swaps on a step-up outsidetransition, andarchive_steppedandasktouch it too, so where the rung changes has three answersbin/loop_claude/loop/reload.py:47decoded_as: a recursive reflection-based decoder with two sentinels and a duck-typedfits; the comments give the reasonbin/loop_claude/cycle/prompts.py:55review_launches: six conditional sentences; checking one branch needsladder.py'sreview_rungsandreplay_appliesbin/loop_claude/cycle/judge.py:185quotes_checked: the misquote rule splits acrossjudge.pyandescalation_mark.py, andgrounded's fragment matching is subtlebin/comprehension_history.py:164worse_places: one nested comprehension to unroll by handskills/larv-review/SKILL.md:84"Re-entry": a table plus exceptions across 🔴, 🟡, the cap and round 2
Mid A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: five near-synonymous rung helpers held at once, and one docstring sentence packs five rules; resolved only acrossescalation_mark.py,ladder.pyandroute.pybin/loop_claude/cycle/run.py:169ask:step_upthreads through three states, and which route each archived line names matters only forrouting_figures/measure.py'srung_atbin/loop_claude/loop/reload.py:47decoded_as: a recursive decoder with two sentinels and duck-typed hooks thatRoutedefines far awaybin/rank_todo.py:428place: recursive with loop detection, in the largest module, which keeps two table shapes side by sidebin/chunk_pick.py:109next_release:Upcoming.landedisbool | Failure, and "landed" has four meanings across modulesbin/loop_claude/cycle/prompts.py:55review_launches: six conditional sentence kinds coupled to the routing rulesbin/comprehension_history.py:159scored: two look-alike predicates that differ on history formats named only by PR numbersbin/loop_claude/loop/main.py:1module docstring: about 120 lines, every sentence a rule, restated again by the units
Maintainability senior, hardest first:
bin/loop_claude/routing.py:76after_unfinished: four rules in one pass over four near-synonym callees; the docstring lists them, but they still had to be simulated by handbin/loop_claude/cycle/run.py:56run_cycle:cycle.routeis replaced inside the loop on a step-up, and a StepUp pair rides through the states so each archive line names the right rungbin/loop_claude/loop/reload.py:47decoded_as: a recursive decoder with sentinels and a duck-typedfits; nothing at the call site says which LoopRun fields survive a reloadbin/loop_claude/cycle/prompts.py:55review_launches: sentences branch on ladder positions, so checking one means computing rungs by handbin/routing_figures/measure.py:104measured: nine locals, timestamps matched as strings cut to the second, and price rescaling on one screenbin/rank_todo.py:332weigh: the OLD_SHAPE compatibility is spread across many functions rather than cut at the boundarybin/comprehension_history.py:159scored: differs fromnewest_scoredonly by a carve-out dated by a PR numberbin/loop_claude/cycle/judge.py:185quotes_checked: three quotes checked against three texts, each misquote with its own outcome; the exit-1 path takes four files
Inherited architect, would restructure first:
README.md:129module map: names bin/'s 18 root modules only as a group, at a level of about 50 entriesbin/round_titles.py:8CAP: the round cap lives here and inbin/rule_hook/round_cap.py:8, not in larv_review/, and nothing says it counts commit titles rather than launchesbin/loop_claude/routing.py:76after_unfinished: step-up and escalation logic spans about eight files, fromcycle/run.pytoloop/main.py'sstepped_routebin/loop_claude/running_loops.py:1module docstring: the inbox file protocol sits inside loop_claude though loop-inbox imports itbin/rule_hook/git.py:7git: shadowsbin/git_repo.py:26git, andreviewer_rung.pyimports and calls bothbin/rung_replay/search.py:17Item: TypeVars namedItemandReplycollide withtodo_table.Itemand judge.py'sReplybin/branch_chunks/cut.py:22Commit: a secondCommitdataclass with other fields thanbin/git_command.py:10's
2026-10-29T20:02:43Z, PR #397 rerun at 1723469
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 5.5 | 6 |
| Mid A | 7 | 6 | 6.5 | 6.5 | 6.5 |
| Maintainability senior | 6.5 | 5 | 6 | 5 | 5.5 |
| Inherited architect | 6.5 | 6 | 5.5 | 7 | 6 |
| Mean | 6.75 | 5.63 | 6.00 | 6.00 | 6.00 |
Junior A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: several conditions combine and hand off to six near-synonym rung helpers; the why sits behinddocs/decisions.mdbin/loop_claude/loop/reload.py:47decoded_as: a recursive decoder with sentinels and an unwrittenfitsconvention, followed only by tracing one call by handbin/routing_figures/measure.py:104measured: many jobs in one function, coupled tocostedand a long comprehension inrenderbin/comprehension_history.py:130newest_scored: a Markdown file parsed as a database, with back-compat notes dated by PR numbersbin/chunk_pick.py:166named_releases: "release" is split acrosschunk_pick,todo_tableandlarv_review/plan.pybin/rank_todo.py:428place: recursive, with OLD_SHAPE threaded through a 658-line file whose help text is the format specbin/loop_claude/cycle/judge.py:185quotes_checked: quote checks layered across three functions andescalation_mark.pybin/rule_hook/repeat_read.py:72shown_lines: timestamps compared as text, and Read and Bash reads treated differently
Mid A, hardest first:
bin/loop_claude/routing.py:76after_unfinished: folds five rules into one function, and the cap it enforces is spread across five filesbin/loop_claude/loop/reload.py:47decoded_as: a hand-written decoder with two sentinels, a union rule for RETIRED and afitshook found bygetattr; three passes to followbin/rank_todo.py:568score: two table shapes run through scoring, weighing and validation, and nothing says which is leavingbin/loop_claude/cycle/prompts.py:55review_launches: six sentence groups from oneReviewChoice, with a replay branch thatladder.py's emptyREPLAYEDleaves without databin/loop_claude/waiting.py:212claim_talk: the/talkhandshake runs between two programs through files and a flock, with deadlines and a compensatingClaudeEndedbin/rule_hook/repeat_read.py:72shown_lines: pairs transcript tool calls with results by id and handles truncation markers of a transcript format written down nowherebin/chunk_pick.py:109next_release: renamed releases, the history since the last tag and landed work; the rename handling stayed unresolvedbin/shell_command.py:44segments: a heuristic token loop with a redirect-target flag; its docstring lists what it doesn't parse
Maintainability senior, hardest first:
bin/loop_claude/routing.py:76after_unfinished: combines the unfinished count, the escalation cap and whether a step already happened, with inputs built incycle/run.py; three files open at oncebin/loop_claude/cycle/prompts.py:55review_launches: about six conditional sentence kinds, with no example output beside the functionbin/loop_claude/loop/main.py:173run_loop: the line rebindingpaths.waitingtolisted.folderquietly points every later question at another folder, explained only atListed.folderbin/model_routing/ladder.py:97review_rungs: the Markdown-only exception that lowers a reviewer lives only inbin/rule_hook/reviewer_rung.py:34, andREPLAYEDis emptybin/loop_claude/loop/reload.py:47decoded_as: sentinels, a duck-typedfitsand recursion over type hints; which archived shapes decode needsroute.pytoobin/rank_todo.py:558goal_term: OLD_SHAPE and SHAPE run through nearly every function, selected by the string "W" or "V"bin/loop_claude/waiting.py:212claim_talk: a three-step handshake over three files, resolved only by holding the whole protocolbin/loop_claude/cycle/judge.py:207grounded: a misquote is handled three ways by what was quoted, with no single table of the outcomes
Inherited architect, would restructure first:
README.md:129module map: bin/'s root holds about 64 ungrouped entries, mixing review, plan and infrastructure modulesbin/rule_hook/round_cap.py:8round_cap_breaches: the cap counts round commit titles, while a third reviewer launch is governed only byskills/larv-review/SKILL.md:76"Round cap"bin/round_titles.py:22round_counts: any unknown pass name gets its own counter, so a misspelled pass never reaches the capbin/loop_claude/routing.py:76after_unfinished: routing has four homes, frommodel_routing/toloop/routing_check.pybin/loop_claude/loop/main.py:256after_verdict: the hub of about 25 imports that every routing or judge feature edits, withstepped_routeon the same pathbin/loop_claude/stop.py:1Stop: documented as loop-claude's exit, but rung_replay exits through it and importscycle.retrytoobin/rank_todo.py:71GOAL_STEP: a separable formula evaluator inside a 658-line, 39-function modulebin/chunk_pick.py:1module docstring: besidebin/chunk_preflight/, the names don't say which does what
2026-11-04T22:52:35Z, PR #415 at 37a412b, against 7681c2b
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Shape:
bin/loop_claude/loop/main.py:159run_loop: reads as a short loop ofnext_cycle,run_cycle,closed_cycleandcheck_routing, where the base held routing, preflight, step-up archiving and the unfinished limit beside it; a symptom leads tochunk_start.pyorclose.py - Locality:
bin/loop_claude/cycle/run.py:40Cycle.questions: the questions folder is an explicit field, where the base reboundpaths.waitingout of sight for every later reader - Locality, Shape:
bin/loop_claude/cycle/run.py:191archived_step: both step-up paths archive through one function and oneStepUp, andnext_routereplaces theafter_unfinished/after_rulingpair, denser but in one place - Navigation, Shape:
bin/loop_claude/waiting/talk.py:1talk: the 233-linewaiting.pyis split by role, so a/talkbug opens a 45-line file - Navigation:
bin/release.py:1release: every release pattern has one home, where the base spread them overchunk_pick,todo_tableandgit_repo - Locality:
bin/release_standing.py:177dated_cut: what is cut on an untagged branch is worked out again at four call sites, andLastCut.tag is Nonequietly means it may be cut by a CHANGELOG heading; weaker than the gains above - Shape:
bin/release_standing.py:1release_standing: the renamed module's docstring still opens "The next chunk's pick"
Inherited architect, decided by:
- Navigation, Shape:
bin/loop_claude/loop/main.py:186next_cycle:run_loopreads as the cycle's outline, with a cycle's start inchunk_start.pyand its close inclose.py, each named for its phase, where the base held about 14 functions in 338 lines - Locality:
bin/loop_claude/loop/main.py:166run_loop: passesquestions=listed.folderexplicitly, where the base reboundpaths.waitingfor every later call - Locality:
bin/loop_claude/cycle/run.py:141rule: archives every interrupt or quoted answer once, as the judge rules, withstates.to_the_humanthe one rule both the transition and the archive read, where the base archived in two places throughran_on - Locality, Shape:
bin/loop_claude/hand_over.py:23hand_over: opening claude has one side effect, no longer a hidden write of~/.claude.jsonthroughtrust.py - Navigation, Shape:
bin/release.py:1release: every release pattern lives in one module, andReleasereplaces the hand-written tuple - Navigation, Shape:
bin/loop_claude/waiting/__init__.py:10waiting: one 234-line module is four of 50 to 80 lines, each a role, at the cost of more imports in its two callers - Locality:
bin/rule_hook/main.py:75breaches_of: the repeat-read helpers moved beside their rule inrepeat_read.py, somain.pyonly dispatches - Locality:
bin/release_standing.py:177dated_cut: whether a CHANGELOG heading counts as the cut is recomputed at four call sites and not kept inLastCut; worse here, outweighed by the places above, and where it rots first - Shape:
bin/loop_claude/routing.py:84next_route: fewer names to learn, a denser body to trace by hand; on balance the same
2026-11-08T18:43:52Z, PR #428 at 4ab3648, against 6e0ef22
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | better | same | better |
| Inherited architect | same | better | better | same | better |
Mid A, decided by:
- Locality, Shape:
bin/comprehension_history.py:127entries: parses each entry once into vote_rows, score_rows, ruled_listings and scored_listings, where the older Entry kept a raw body that every consumer re-parsed under order rules ("the votes table comes first", "the first row wins") - Shape:
bin/comprehension_history.py:168newest_commit_runs: says what it returns, where the oldernewest_scoredpacked the same logic into one conditional comprehension, andscoring_runswith its caveat is gone - Locality:
bin/rule_hook/rules.py:161unlimited_container: asksfinds_this_dockerof the module beside run-heavy's docker, where the older version did index arithmetic on the words before docker and kept its ownresets_path - Shape:
bin/model_routing/route.py:81route_from: reads one format, where the older version also kept a legacy regex andlegacy_fields - Shape:
bin/deploy_release/main.py:44main: a thin dispatcher overdeploy.py,cut.pyandforge.py, each cohesive
Inherited architect, decided by:
- Locality, Shape:
bin/rule_hook/rules.py:166unlimited_container: the PATH reasoning now sits infinds_this_dockerandresets_pathinbin/run_heavy/capped_docker.py:39next to the docker it describes, andLaunchcarrieswordandbefore - Locality, Shape:
bin/comprehension_history.py:127entries: parses each entry once; worse_places readslisting.seatinstead of splitting a label, and the tables stay separate so no ordering trap remains - Shape:
bin/model_routing/route.py:81route_from: the legacy worker-line format is gone, so one format is left to read and keep in sync - Shape:
bin/review_replay/main.py:100main: the oldanchorvariable is renamedproject, matching the project's-rung vocabulary
2026-11-09T11:09:33Z, PR #453 at f213dfb, against 0b77d12
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Shape:
README.md:133bin/import-root paragraph: the base left 15 shared modules loose beside the launchers; the head puts them in named packages that the README lists with who imports them. - Locality, Shape:
bin/release_plan/ranking.py:60rank_file: the base'srank_todo.pywas imported as a library by two tools; the head's script is a thin shell, and item 274'sweight_ofatranking.py:469keeps its rule in one function. - Shape, Navigation:
bin/routing_figures/chunk.py:62chunk_figures: the base's one 367-line module is split into files named for what they hold. - Self-sufficiency, Locality:
bin/claude_code/entries.py:18Entry: the head parses each transcript line into a typed dataclass once, where the base passed raw dicts. - Locality:
bin/loop_claude/cycle/judge.py:117Brief: five positional arguments travel as one documented tuple, and the inline checks are named steps.
Inherited architect, decided by:
- Navigation, Shape:
bin/rank_todo.py:88main: the 660-line script is a launcher and the logic lives inbin/release_plan/ranking.py:60rank_file, whose callers the README names. - Navigation:
README.md:132bin/import-root paragraph: 32 flat modules became named packages with a map of who reads each. - Shape, Locality:
bin/routing_figures/chunk.py:62chunk_figures: four concerns split into files with typed entries. - Locality:
bin/routing_figures/state.py:45due: the archive-root state has one owner. - Locality:
bin/release_plan/ranking.py:469weight_of: item 274's rule for a maintain point's weight sits in one function instead of two places.
2026-11-10T21:39:35Z, PR #467 at 05539e2, against 98b7bfd
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Locality:
bin/loop_state/__init__.py:1loop_state: the files loop-claude and loop-inbox share have their own package, whose docstring says a change there changes both tools - Navigation, Shape:
bin/loop_effects/effects.py:1Effects: the replay tools and the trials driver importloop_effects.effects, not a private-lookingloop_claude.effects - Locality, Shape:
bin/release_plan/ranking.py:1ranking: blocker order, formula parser and disk I/O moved to named files, so one can change without the score weighting in view - Locality, Shape:
bin/loop_claude/loop/this_loop.py:12Loop: seven to nine positional arguments, two of them swappable strings, became one frozen value - Locality:
bin/loop_claude/cycle/run.py:93archived: five copies of the archive call became one named helper - Navigation:
README.md:116"bin/ map": the map follows the tools directly and testing has its own heading
Inherited architect, decided by:
- Navigation, Locality:
bin/loop_state/__init__.py:1loop_state: an inbox symptom now lands in a package named for the shared state, not seven modules insideloop_claude/ - Shape, Locality:
bin/release_plan/ranking.py:56rank: the 550-line module's I/O, formula and blocker order each have an obvious file - Locality, Shape:
bin/loop_claude/routing.py:84next_route: each verdict kind has its own function, andstepsandbelow_capname the two conditions - Locality:
bin/model_routing/route.py:88route_of: the archive line and the reload hand-over decode a Route through one decoder - Shape:
bin/loop_claude/loop/main.py:66next_cycle: signatures of seven or eight positional parameters shrink to(loop_run, loop)
2026-11-10T23:11:34Z, PR #473 at 13d1550
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6 | 6 |
| Mid A | 7 | 5.5 | 6 | 6.5 | 6 |
| Maintainability senior | 7 | 5.5 | 6 | 6 | 6 |
| Inherited architect | 7 | 5.5 | 6 | 7 | 6 |
| Mean | 7.00 | 5.50 | 6.00 | 6.38 | 6.00 |
Junior A, hardest first:
- Locality, Shape:
bin/loop_claude/routing.py:93unfinished_route: the floor rung, the escalation mark, the unfinished count and the cap are held at once, with escalation_mark.py open beside it - Locality, Shape:
bin/larv_review/plan.py:66plan_reasons: eight failure branches mix git reads, file reads and logic, and a parameter nameddefaultreceives a merge-base commit - Locality:
bin/larv_review/owed.py:37shortfall: whether one or two runs are owed hangs on the wording of an English sentence that SKILL.md also keys on - Shape:
bin/loop_claude/cycle/run.py:59run_cycle: four near-identical archive helpers and a route replaced partway through the loop - Shape:
bin/release_plan/ranking.py:56rank: nearly every function forks on the table's two shapes - Self-sufficiency:
skills/larv-review/SKILL.md:147"Scoring run": dense rule prose with many cross-references, three sections to rebuild one rule
Mid A, hardest first:
- Locality, Shape:
bin/larv_review/plan.py:66plan_reasons:defaultholds a merge-base commit on one call path andaftermeans the checkout's text, then its items - Locality:
bin/loop_claude/routing.py:93unfinished_route: six concepts meet in about twelve lines and need four modules open beside it - Locality:
bin/larv_review/owed.py:37shortfall: a rule keyed on the text of a sentence that plan.py builds, andscoring_runsdefines a scoring run differently fromEntry.scored - Locality:
bin/loop_claude/cycle/judge.py:211quotes_checked: a verdict is rewritten four times, half the behaviour in prompt prose and half in code - Shape:
bin/release_plan/ranking.py:304score: every function branches on the old and the current table shape
Maintainability senior, hardest first:
- Locality, Shape:
bin/larv_review/plan.py:66plan_reasons: a merge-base commit enters the parameter named for the default branch, and its error text would print the commit as a branch - Locality:
bin/larv_review/owed.py:37shortfall: the one-run-or-two rule is spread over three files and a skill document - Locality, Shape:
bin/loop_claude/routing.py:93unfinished_route: the six inputs interlock and a wrong answer runs the worker on the wrong rung without failing - Locality:
bin/model_routing/route.py:88route_of: a route has two serializations, a flat archive line and a nested handover - Self-sufficiency:
bin/larv_review/plan.py:66plan_reasons: its docstring doesn't say the first argument may be a commit
Inherited architect, would restructure first:
- Navigation, Shape:
bin/larv_review/owed.py:10owed_scoring_run: the name gives the merge gate's question, not the module's subject, and the scoring-run logic sits in three files whose imports read as circular - Locality:
bin/larv_review/plan.py:14EMPTIED_WORDS: prose and code share an exact string that SKILL.md keys on - Locality:
bin/loop_claude/routing.py:93unfinished_route: routing logic spreads over about eight places, so a new trigger touches five files in three packages - Shape:
bin/rule_hook/rules.py:25RULES: rules sit inline in one module while five others have a module each - Navigation:
bin/README.md:116 "bin/": 56 entries in one flat list, the worst fan-out level
2026-11-11T12:55:06Z, PR #475 at 52da15e
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6 | 6 |
| Mid A | 7.5 | 6 | 6.5 | 6.5 | 6.5 |
| Maintainability senior | 7 | 6 | 6.5 | 6 | 6.5 |
| Inherited architect | 7 | 6.5 | 6 | 7 | 6.5 |
| Mean | 7.13 | 6.00 | 6.25 | 6.38 | 6.38 |
Junior A, hardest first:
- Locality, Shape:
bin/loop_claude/routing.py:93unfinished_route: the floor, the escalation cap, the escalation mark, the unfinished count and the taken-in E + I are held at once - Locality:
bin/loop_claude/cycle/judge.py:211quotes_checked: the judge's contract is split over prose constants, schema descriptions and post-hoc checks that change a verdict's type - Locality:
bin/loop_claude/cycle/run.py:146rule: archive order carries meaning acrossrun_cycleandloop/close.py:18closed_cycle - Shape:
bin/release_plan/release_standing.py:145next_release: tag, changelog, before-first-release and renamed branches meet in theis_earliestexpression - Shape:
bin/release_plan/ranking.py:56rank:OLD_SHAPEandSHAPEforks run through nearly every function - Self-sufficiency:
bin/loop_claude/cycle/prompts.py:59review_launches: the worker's prompt is built from about ten conditional sentence builders - Locality:
skills/larv-review/SKILL.md:76"Round cap": the cap is enforced in four units in two places
Mid A, hardest first:
- Locality, Shape:
bin/loop_claude/routing.py:93unfinished_route: five things applied in a fixed order across about six helpers; the reason for the floor and cap sits in the decision log - Locality:
bin/loop_claude/cycle/run.py:146rule: the order of ruling, rejection and step-up archive writes is stated piecemeal - Shape:
bin/release_plan/ranking.py:315weight_of: old-shape branches double the module, and the constants are magic until rules/todo-items.md is open - Shape:
bin/release_plan/release_standing.py:145next_release: a three-way boolean on line 167 with no comment for theNoneandnothing_cutcase - Locality:
bin/rule_hook/repeat_read.py:119shown_lines: rebuilds which lines are in the agent's context from transcript entries - Locality:
bin/loop_claude/cycle/judge.py:211quotes_checked: half of the judge's rules are English constants edited in step with Python - Locality:
bin/loop_claude/cycle/prompts.py:119escalation: about 20 sentence builders restate the routing logic
Maintainability senior, hardest first:
- Shape, Locality:
bin/release_plan/release_standing.py:145next_release: four fallbacks held at once, andis_earliesttook three reads - Locality:
bin/loop_claude/routing.py:93unfinished_route: the escalation mark's meaning lives inescalation_mark.py, which has to stay open beside it - Locality:
bin/loop_claude/cycle/judge.py:211quotes_checked: a bad quote silently changes the verdict's type - Shape:
bin/release_plan/ranking.py:304score: every path forks onSHAPEversusOLD_SHAPE - Locality:
bin/rule_hook/reviewer_rung.py:40skill_breaches: three cases rest on assumptions made in another function - Locality:
bin/loop_claude/cycle/run.py:146rule: order-dependent archive writes that nothing enforces - Shape:
bin/review_rules/comprehension_history.py:127entries: a line-by-line parser that tolerates two legacy formats - Self-sufficiency:
skills/larv-review/SKILL.md:76"Round cap": dense cross-referencing rules, with the enforcing code in other modules
Inherited architect, would restructure first:
- Navigation, Shape:
README.md:116bin/: 55 entries at one level, about 31 packages, the worst fan-out level - Locality:
bin/larv_review/reviewer_header.py:49unanswered_round: the round cap counts answer commits, so a worker that relaunches without committing gets the same round number, and the rule is split over SKILL.md and three modules - Locality, Shape:
bin/loop_claude/routing.py:93unfinished_route: seven counters and limits across five files, with field names frozen by the reload handover - Locality:
bin/chunk_pr/main.py:42forge: the forge helpers live inchunk_preflightanddeploy_release, so a second forge needs edits in four packages - Shape:
bin/records/__init__.py:1records: one name for three concepts, ascutandPlanalso name several - Locality:
bin/loop_claude/cycle/run.py:86archived: the archive helpers thatloop/close.pyimports sit in the cycle runner
2026-11-12T18:40:20Z, PR #510 at 62faf32
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 6.5 | 5 | 5.5 | 4.5 | 5.5 |
| Mid A | 7 | 6 | 6.5 | 5 | 6 |
| Maintainability senior | 7 | 6 | 6 | 5 | 6 |
| Inherited architect | 6.5 | 7.5 | 6 | 5.5 | 6.5 |
| Mean | 6.75 | 6.13 | 6.00 | 5.00 | 6.00 |
Junior A, hardest first:
- Locality, Shape:
bin/loop_claude/routing.py:84next_route: item rung, floor, cap, escalation mark and unfinished count spread over four modules, and the reasons sit in the decision log - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: each check silently demotes the verdict's type, andINTERRUPT_RULESis policy held as a prompt string - Locality, Shape:
bin/loop_claude/cycle/run.py:154rule: one function reads, judges, archives and steps up in an order a comment alone states - Self-sufficiency:
skills/larv-review/SKILL.md:35"Passes": about 200 lines of nested conditions whose terms are defined at the end of CLAUDE.md - Shape:
bin/larv_review/main.py:173main: a 150-line docstring doubles as the spec of about 12 subcommands in one elif chain - Locality, Shape:
bin/release_plan/ranking.py:304score: two table shapes and a hidden comprehension-history input - Locality:
bin/larv_review/scoring_setup.py:156prepare_scoring: code parses two Markdown files, so a prose edit can break it
Mid A, hardest first:
- Locality, Shape:
bin/loop_claude/routing.py:84next_route: rung, floor, cap, mark and unfinished count feed one choice across four files - Self-sufficiency:
skills/larv-review/SKILL.md:76"Round cap": a rule engine in English whose numbers are repeated in code, CLAUDE.md and--help - Locality:
bin/loop_claude/cycle/prompts.py:61review_launches: the worker's text is assembled from rung choices and has to be run in the head - Shape:
bin/command_lines/shell_command.py:46segments: a hand-written shell tokenizer whose regexes stay opaque - Locality:
bin/rule_hook/repeat_read.py:119shown_lines: depends on Claude Code's transcript format - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: five quote checks with different failure modes - Shape:
bin/release_plan/release_standing.py:145next_release: many tag, rename and changelog cases in a six-way union - Shape:
reviewer-merge-trials/driver.py:1driver: one-off scripts that break the repo's tab and naming conventions
Maintainability senior, hardest first:
- Locality, Shape:
bin/release_plan/release_standing.py:145next_release: six cases held at once,Cutagainstcut, and a compoundis_earliestcondition - Shape:
bin/release_plan/ranking.py:304score:SHAPEagainstOLD_SHAPEthreaded through nearly every function - Locality:
bin/loop_claude/routing.py:84next_route: rung arithmetic split withescalation_mark.py - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: each layer silently downgrades the verdict - Locality, Shape:
bin/loop_claude/cycle/run.py:154rule: order-dependent steps and a 17-fieldCycle - Locality:
bin/larv_review/main.py:173main: the scoring-run rule is split overplan.py, unopened siblings and a wording match with SKILL.md - Self-sufficiency:
skills/larv-review/SKILL.md:76"Round cap": mechanical rules duplicated in three code places, so no copy is authoritative
Inherited architect, would restructure first:
- Navigation, Shape:
README.md:96bin/: about 100 entries at one level, with the README's two lists as the only map - Locality:
bin/review_rules/round_titles.py:1ROUND: the commit-title protocol is parsed in five tools and written in SKILL.md - Shape:
bin/release_plan/release_standing.py:41Next:Next,CutandLastCuteach carry more than one meaning - Shape:
bin/records/__init__.py:1records: one name for two unrelated concepts, and "plan" and "ruling" also name several - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: the judge's rules are strings coupled to CLAUDE.md's text - Shape:
bin/rule_hook/rules.py:127check: round-cap rules sit beside unrelated pin rules in 13 rules - Self-sufficiency:
bin/loop_claude/routing.py:84next_route: the reasons cite decision titles the reader cannot see
2026-11-12T18:43:58Z, PR #510 rerun at fba48f3
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 6.5 | 5 | 5.5 | 4.5 | 5.5 |
| Mid A | 7 | 5.5 | 6.5 | 5 | 6 |
| Maintainability senior | 7 | 5.5 | 6 | 5 | 6 |
| Inherited architect | 7 | 6.5 | 6 | 5.5 | 6 |
| Mean | 6.88 | 5.63 | 6.00 | 5.00 | 5.88 |
Junior A, hardest first:
- Self-sufficiency, Locality:
skills/larv-review/SKILL.md:80"Round cap": nested rules that point at CLAUDE.md,standards/,rules/and decision titles, so three documents stay open at once - Locality, Shape:
bin/loop_claude/routing.py:93unfinished_route: floor, cap, mark and unfinished count interact acrossRoute,Rungandescalation_mark - Locality:
bin/release_plan/release_standing.py:145next_release: no tag, dated headings, renames andnamed_releasesmeet in one function - Self-sufficiency, Shape:
bin/release_plan/ranking.py:271goal_weights: paragraph-long docstrings carry the rules, and the constants cite a decision log - Locality, Shape:
bin/loop_claude/cycle/run.py:154rule: one function with four near-identically named archive helpers beside it - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: ellipsis matching and a policy that only a misquoted evidence fails - Shape:
bin/review_replay/main.py:225code_of: guesses which code was reviewed from archive times and prompt text
Mid A, hardest first:
- Locality, Shape:
bin/release_plan/ranking.py:56rank: every helper forks onSHAPEversusOLD_SHAPE - Locality, Shape:
bin/release_plan/release_standing.py:145next_release: three inputs and sixPicktypes - Locality:
bin/loop_claude/routing.py:93unfinished_route: the rule that a rung never falls spans several functions - Locality:
bin/loop_claude/cycle/run.py:67run_cycle: replacescycle.routepartway through its loop, out of sight oftransition - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: each check silently demotes the verdict's type - Locality:
bin/review_rules/comprehension_history.py:240listings: a Markdown format parsed by regex and shared withlarv_reviewandrank_todo - Locality:
bin/rule_hook/repeat_read.py:119shown_lines: rebuilds the agent's context from Claude Code transcripts - Self-sufficiency:
skills/larv-review/SKILL.md:80"Round cap": needsdocs/decisions.md,rules/*andstandards/*open beside it
Maintainability senior, hardest first:
- Self-sufficiency, Locality:
skills/larv-review/SKILL.md:35"Passes": a process spec in which nothing marks which sentences code enforces, and where - Locality:
bin/loop_claude/routing.py:93unfinished_route: rung state shared withescalation_mark.pyandmodel_routing/ladder.py - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: prompt strings are behaviour with no test-shaped boundary - Locality, Shape:
bin/loop_claude/cycle/run.py:154rule: the route lives inCycleoutside the state machine, and archive order is only commented - Shape:
bin/release_plan/ranking.py:304score:OLD_SHAPEexists only to read old history, and ranking reads a documentation file as data - Shape:
bin/release_plan/release_standing.py:145next_release: sixPickvariants and layered tag cases - Locality:
bin/larv_review/scoring_setup.py:65score_bullets: parses SKILL.md's prose bullets and fails on a count change, so a prose edit is load-bearing
Inherited architect, would restructure first:
- Locality:
bin/review_rules/round_titles.py:59past_cap: the round cap lives in four code places and in SKILL.md §3, and the hook counts commit titles, not launches - Shape, Locality:
bin/release_plan/ranking.py:56rank:OLD_SHAPEbranches in six functions, and the module also parses goals and checks decision citations - Locality:
bin/release_plan/release_standing.py:145next_release: release standing is also read bylarv_review/plan.py,chunk_preflightandloop_claude/routing.py - Navigation, Shape:
README.md:96bin/: about 85 entries at one level, with the launcher-to-package rule as the only map - Shape:
bin/larv_review/main.py:173main: a 145-line docstring doubles as help beside an elif chain - Shape:
bin/rule_hook/rules.py:14RuleName: a Literal that repeats the keys ofRULES, asReviewedPassrepeatsPASS_AGENTS - Self-sufficiency:
bin/loop_claude/loop/chunk_start.py:83fetched_plan: a comment cites a plan item that is deleted when finished
2026-11-12T19:51:29Z, PR #511 at 9a260d2, against 913accf
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | better | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Shape, Overall:
bin/review_replay/main.py:75main: the file is now CLI only, and each job sits in its own module under a docstring stating it; at913accfone 359-line file mixed them - Locality, Shape, Overall:
bin/larv_review/main.py:211COMMANDS: one table row per subcommand feeds dispatch, usage and--help, where913accfhad a 12-arm if/elif chain and a hand-written usage string - Locality, Navigation:
bin/transcript_archive/outcomes.py:63spent_block: the Spent block's writer now sits beside its marker and kinds, not in another package - Locality:
bin/claude_code/entries.py:148messages: one reader groups assistant turns by message, where three callers each deduplicated by hand with different keep-first and keep-last rules - Shape, Self-sufficiency:
bin/model_routing/prices.py:23PRICES: the docstring states both uses of the price list and the comment says prices are archived as in force - Self-sufficiency, Locality:
bin/larv_review/scoring_setup.py:52score_bullets: picks anchors and dimensions by how each bullet opens, and the docstring says so;913accftook them by position and count - Shape:
bin/review_replay/omit_comparison.py:12compared: returns the report and leaves printing tomain, where913accfprinted, returned an exit code and needed a cast
Inherited architect, decided by:
- Navigation, Shape, Overall:
bin/review_replay/replay.py:1replay: a failed clone or bad match leads to one file whose docstring names that job;main.pyshrank from 359 to 114 lines of wiring - Locality, Shape:
bin/claude_code/entries.py:148messages: one reader groups a message's entries, and the keep-first versus keep-last trap the old comment warned about is gone - Navigation, Locality:
bin/larv_review/main.py:211COMMANDS: each subcommand's spelling, argument check and action sit in one row that usage and--helpare built from - Locality:
bin/transcript_archive/outcomes.py:63spent_block: the writer sits beside the marker and parser, and the loop no longer depends on the measuring package for an archive format - Shape:
bin/model_routing/prices.py:1prices: the price list moved out ofrouting_figuresonce it gained a second reader, and its docstring states both uses
2026-11-12T20:22:47Z, PR #513 at d5a9576, against 0a26bcb
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | better | same | same | better |
| Inherited architect | same | better | better | same | better |
Mid A, decided by:
- Locality, Overall:
bin/larv_review/scoring.py:10is_floor_item: the rule for what makes a plan item a floor item lives in one named predicate that both callers use, where0a26bcbwrote the same test out twice - Locality:
bin/larv_review/plan.py:28scoring_run_due: readsis_floor_item(item)and printsa floor item, so the trigger needs no knowledge of the tagline's format
Inherited architect, decided by:
- Locality, Shape, Overall:
bin/larv_review/scoring.py:10is_floor_item: one named predicate that bothscoring_run_dueandfloor_stepcall, where each spelled its own substring test at0a26bcb - Shape:
bin/larv_review/scoring.py:7FLOOR_ITEM: the constant is now the tagline's opening phrase and matched as one, where it was a bare substring that any mention matched - Locality:
bin/larv_review/plan.py:28scoring_run_due: the reason line reads off the predicate, so the function no longer carries tagline matching
2026-11-12T20:40:16Z, PR #514 at bbd5630, against e705fd8
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | same | same | same | same |
| Inherited architect | same | same | same | same | same |
2026-11-13T01:19:17Z, PR #520 at 8f03747
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 5.5 | 6 |
| Mid A | 6.5 | 5.5 | 6 | 5 | 6 |
| Maintainability senior | 7 | 6 | 6 | 6 | 6.5 |
| Inherited architect | 7 | 6 | 6 | 6.5 | 6.5 |
| Mean | 6.88 | 5.75 | 6.00 | 5.75 | 6.25 |
Junior A, hardest first:
- Locality, Self-sufficiency:
bin/loop_claude/routing.py:93unfinished_route: five notions of rung plus the escalation mark andUNFINISHED_PER_STEP, with each step's reason in the decision log - Locality:
bin/loop_claude/cycle/run.py:154rule: stamps, annotates, steps up and archives in an order only the code states, withclose.pyarchiving again - Locality, Shape:
bin/release_plan/ranking.py:304score: every function forks onSHAPEversusOLD_SHAPE, with no word on when the old one goes - Shape:
bin/release_plan/release_standing.py:168next_release: sixPicktypes and| Failureor| Nonefields inLastCut - Shape:
bin/rule_hook/todo_check.py:22todo_breaches: three config paths, and why the checkout's own script runs only undersame_repostays unclear - Shape:
bin/loop_claude/loop/reload.py:220decoded_route: one dense guard, and renaming aLoopRunfield stops a running loop - Self-sufficiency:
skills/larv-review/SKILL.md:82"Round cap": needsCLAUDE.md,review-findings.mdand the rules files open at once
Mid A, hardest first:
- Locality, Self-sufficiency:
bin/loop_claude/routing.py:93unfinished_route: five rung notions across four helpers and the escalation mark, the cap's reason only in the decision log - Shape, Locality:
bin/rule_hook/todo_check.py:22todo_breaches: four paths,held()andlocated()answering similar questions differently, andchangedhashing a blob by hand - Shape:
bin/release_plan/release_standing.py:168next_release: three-wayLastCutfields whoseNonemeans different things per field - Locality:
bin/release_plan/ranking.py:304score: two table shapes branch through scoring, and the weights' reasons sit elsewhere - Locality:
bin/loop_claude/cycle/judge.py:196quotes_checked: six checks each turn the verdict into something different - Shape:
bin/larv_review/plan.py:66plan_reasons: several failing reads threaded through every branch, andids=Nonemeans something only the callers say - Locality:
bin/loop_claude/cycle/run.py:154rule: a fixed order nothing enforces, and an 18-fieldCyclepassed everywhere - Self-sufficiency:
bin/larv_review/main.py:1help: about 150 lines of rules a reader needs before any handler
Maintainability senior, hardest first:
- Locality, Self-sufficiency:
bin/loop_claude/routing.py:93unfinished_route: six things held at once, and why the rules combine sits behind the decision log - Shape:
bin/rule_hook/todo_check.py:42folder_breaches: four cases and a second fork incheck_failures, andheldneither says what it tells nor thatconfig.topis the main checkout - Shape:
bin/rule_hook/rules.py:82todo-rank-check: one help sentence carries three conditions joined by "and" - Locality:
bin/loop_claude/cycle/run.py:154rule: archives twice in a set order, andstepped_tochangescycle.routeoutside the state machine - Locality, Self-sufficiency:
bin/release_plan/ranking.py:160weigh: two table formats, and constants whose reason is only "docs/decisions.md records the choice" - Shape:
bin/rule_hook/main.py:116bash_breaches: a 200-character return line joining four kinds of breach - Locality:
bin/larv_review/main.py:1help: the only full statement ofneeds-scoring-runandscoring-run-owed, implemented in other modules
Inherited architect, would restructure first:
- Locality, Shape:
bin/chunk_preflight/choices.py:80forge_of: forge code split acrosschunk_preflightanddeploy_release, which three tools reach into - Navigation:
bin/records/__init__.py:1records: file utilities share a name with the judge's records inloop_claude/project/records.py - Shape:
README.md:3dated experiments: five research and outreach folders at the root beside the working areas - Navigation, Shape:
README.md:131bin/: 66 entries at one level, navigable only through the README's prose map - Locality:
bin/loop_claude/cycle/run.py:154rule:rule,closeandkeep_rejectioneach archive and map a Failure toStop(1) - Self-sufficiency:
bin/rule_hook/round_cap.py:26earlier_titles: fails open when git cannot list the branch's titles, so a mislabelled round passes
2026-11-13T01:23:02Z, PR #520 rerun at 8f03747
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 6.5 | 5 | 5.5 | 5 | 5.5 |
| Mid A | 7 | 5.5 | 6 | 5.5 | 6 |
| Maintainability senior | 7 | 6 | 6 | 6 | 6.5 |
| Inherited architect | 6.5 | 6 | 5.5 | 6 | 6 |
| Mean | 6.75 | 5.63 | 5.75 | 5.63 | 6.00 |
Junior A, hardest first:
- Shape:
bin/release_plan/release_standing.py:168next_release: a before-first-release constant, cut and named releases, andstr | None | Failurefields tracked to where each is checked - Locality, Self-sufficiency:
bin/loop_claude/routing.py:93unfinished_route: the escalation rule split across four files, five ideas in a few lines, and the reasons in the decision log - Shape:
bin/rule_hook/todo_check.py:22todo_breaches: three paths, the checkout's own script only undersame_repo, and a blob hashed by hand - Locality:
bin/loop_claude/cycle/run.py:67run_cycle: changescycle.routeaway from where the 17-fieldCycleis built - Locality:
bin/larv_review/owed.py:46shortfall: logic hangs on the exact wording of a sentence SKILL.md also depends on - Shape:
bin/rule_hook/committed_files.py:14committed_file_breaches: a reversed-dict dedupe whose choice of copy no comment explains - Self-sufficiency:
skills/larv-review/SKILL.md:82"Round cap": every sentence points to another rule, and the code read easier than the prose - Shape:
bin/release_plan/ranking.py:315weight_of: old and new table shapes in five functions, and a docstring packing three rules into one sentence
Mid A, hardest first:
- Locality, Self-sufficiency:
bin/loop_claude/routing.py:93unfinished_route: five things held for one step-up, and why the rules interact lives in the decision log - Locality:
bin/loop_claude/cycle/run.py:154rule: archives twice in an order that matters, repeated acrossrun.pyandclose.py, and nothing guards it - Self-sufficiency:
bin/release_plan/release_standing.py:168next_release:BEFORE_FIRST_RELEASEand the earliest test needCLAUDE.md→ "Working loop" open beside them - Shape:
bin/release_plan/ranking.py:56rank: each function reads as two algorithms interleaved, old shape and new - Self-sufficiency:
bin/larv_review/plan.py:66plan_reasons: the 155-line help and the code each state part of when a scoring run is due - Self-sufficiency:
bin/loop_claude/cycle/judge.py:21INTERRUPT_RULES: behaviour held in prompt text and schema descriptions the code can't explain - Shape:
bin/rule_hook/todo_check.py:22todo_breaches: three branches, and a subprocess path I would not have guessed - Self-sufficiency:
skills/larv-review/SKILL.md:82"Round cap": dense rules pointing at each other, some at files out of reach
Maintainability senior, hardest first:
- Shape, Self-sufficiency:
bin/release_plan/release_standing.py:168next_release:Failurein everyLastCutfield, a sentinel version and anothing_cutflag worked through as truth tables - Locality:
bin/loop_claude/routing.py:93unfinished_route: five helpers and the escalation mark held at once, the rule kept in no one place - Shape:
bin/release_plan/ranking.py:160weigh: the legacy shape doubles every scoring function, and nothing says when it can go - Shape:
bin/rule_hook/committed_files.py:14committed_file_breaches: a double-reverse dedupe with no comment - Locality:
bin/loop_claude/cycle/run.py:154rule: five steps in a row, overlapping archive helpers, and an order that is the trap - Shape:
bin/rule_hook/todo_check.py:42folder_breaches:config.foldertochangedbutconfig.toptocheck_failures, right but a stop to confirm - Shape:
bin/model_routing/route.py:28levels_fit: two rules in one boolean, androute_ofpacks its type checks into one long line - Self-sufficiency:
skills/larv-review/SKILL.md:82"Round cap": dense bullets pointing elsewhere, reread before the code matched
Inherited architect, would restructure first:
- Navigation, Shape:
README.md:131bin/: about 63 entries at one level, with the shared libraries told apart only by README prose - Locality, Shape:
bin/chunk_preflight/choices.py:80forge_of: forge identity and forge calls in two tool packages, so a second forge lands in three - Navigation:
bin/records/__init__.py:1records: a utility package named like the judge's records, and quote semantics inclaude_code/transcript.py:85verbatim - Locality:
bin/loop_claude/cycle/run.py:94archived: archive helpers in the cycle runner, whichloop/close.pyimports - Shape:
README.md:3trial folders: four finished experiments at the root beside the working areas - Locality:
bin/rule_hook/rules.py:61round-cap: the cap's wording spelled in SKILL.md,rules.pyandround_titles.py - Self-sufficiency:
bin/rule_hook/round_cap.py:26earlier_titles: fails open on the unattended path when git cannot list the branch's titles
2026-11-13T02:49:59Z, PR #521 at ade48c57b5, against 6b7392adcb
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Shape, Navigation:
bin/release_plan/goals_file.py:34ranked_goals: the goals-file logic (Void handling, numbering, goal weights) has its own module, which the README names; at the base it sat in a nearly 400-lineranking.py, mixed with table parsing and scoring - Locality, Shape:
bin/loop_claude/project/premises.py:9read_premises: sharesgoals_file.sectionwith the ranking, so a heading rule changes in one place; at the base it repeated the missing, repeated and empty checks - Navigation, Shape:
bin/loop_claude/quotes.py:4verbatim: the judge's quoting rule sits beside its only users; at the base it sat in the general transcript reader, where a reader hunting a misquote rejection would not look - Shape, Navigation:
bin/text_files/markdown_text.py:1text_files: the package name says what it does; at the base it wasrecords, which in this domain means the judge's records and past panels' records - Locality:
bin/larv_review/scoring_setup.py:169prepare_scoring:homeis passed in and reads go through onereadhelper that returns aFailure; against that the function grew longer, andprompt_partsbuilds theseat-prompts.mdpath twice - Locality:
bin/release_plan/plan_file.py:10rank_file: returns its report lines for the caller to print; at the base it printed to stderr itself - Locality:
bin/review_rules/round_titles.py:12OPENING: the round title's spelling sits above the regexes that parse it, so changing the format touches one place
Inherited architect, decided by:
- Navigation, Shape:
bin/release_plan/goals_file.py:24section: the goals file's sections, numbering, void rule and weights have a module named for what it reads; at the base they sat inside a 400-line ranking module, mixed with table parsing - Locality:
bin/loop_claude/project/premises.py:17read_premises: loop-claude's premises and the ranking read## Goalsthrough onesection; at the base each had its own checks, which could drift apart - Shape, Navigation:
README.md:181text_files/: the new name tells the truth about a package of Markdown sections, UTC stamps and atomic writes;records/meant the decision log and the histories in this repo - Shape:
bin/loop_claude/quotes.py:4verbatim: loop-claude's quote-matching policy left the shared transcript parser, which now holds only transcript reading - Locality:
bin/release_plan/plan_file.py:23report_lines: the library returns lines and the caller decides; at the baserank_fileprinted to stderr as a hidden side effect - Locality:
bin/review_rules/review_folder.py:17made_review_folder:homeis passed in from the command table, so each unit's inputs show in its signature; at the basePath.home()was read deep inside - Shape:
bin/larv_review/scoring_setup.py:169prepare_scoring: counts against the change: about 75 lines with a dozen early returns, andprompt_partsbuilds one path twice - Shape:
bin/review_rules/round_titles.py:13OPENING: the round-title spelling is written once next to the regexes that parse it
2026-11-15T21:26:29Z, PR #554 at bcb04e3
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 6.5 | 5 | 5.5 | 5 | 5.5 |
| Mid A | 7 | 6 | 6.5 | 6 | 6.5 |
| Maintainability senior | 7 | 5.5 | 6 | 6 | 6 |
| Inherited architect | 6.5 | 6 | 5.5 | 5.5 | 6 |
| Mean | 6.75 | 5.63 | 5.88 | 5.63 | 6.00 |
Junior A, hardest first:
- Locality, Self-sufficiency:
bin/loop_claude/routing.py:113unfinished_route: five rung notions, two step-up counters and a seven-row verdict table at once, and the reasons in the decision log - Shape:
bin/release_plan/release_standing.py:162next_release: many edge-case branches andstr | None | Failurethreaded through every field ofLastCut - Locality, Shape:
bin/loop_claude/cycle/run.py:151rule: judges, stamps, annotates, steps the rung up and archives, and no single unit owns the archive's order - Self-sufficiency:
bin/loop_claude/cycle/judge.py:22INTERRUPT_RULES: policy partly in prompt text and partly in Python checks, with nothing saying which the code enforces - Locality:
bin/loop_claude/loop/reload.py:95exec_if_edited: an edit to any.pyunderbin/makes a running loop re-exec, and a renamedLoopRunfield breaks the reload - Shape:
bin/model_routing/ladder.py:86review_rung: a three-way rule whose docstring quotes a whole decision title - Self-sufficiency:
bin/larv_review/main.py:1module docstring: twelve subcommands' rules in about 155 lines of prose that cross-reference SKILL.md - Shape:
bin/loop_claude/loop/chunk_start.py:24fetched_todo: nearly the same plan read asfetched_plan, told apart only by a docstring
Mid A, hardest first:
- Shape:
bin/release_plan/release_standing.py:162next_release: a three-state tag,BEFORE_FIRST_RELEASE,nothing_cutandis_earliestheld at once, and comments that explain each branch alone - Locality:
bin/loop_claude/routing.py:113unfinished_route: five rungs, the escalation mark and the unfinished count feed one answer across three files - Locality, Shape:
bin/chunk_pr/main.py:298merged: a while loop around the queue turn, variables bound only on some paths, and no comment stating the loop's invariant - Shape:
bin/chunk_preflight/main.py:128preflight: lines 142-143 pack a conditional re-locatedand an eight-argumentGatheredinto two lines - Shape:
bin/release_plan/ranking.py:48rank: two table shapes branch through the whole module, and the defect rule sits far from where scores are read - Self-sufficiency:
bin/larv_review/plan.py:77plan_reasons: six inputs joined into one answer, and why each matters is in SKILL.md §3 - Shape:
bin/rule_hook/env_file.py:34reaching:andandormixed without parentheses, and four path spellings in a lookaround regex - Self-sufficiency:
bin/loop_claude/cycle/judge.py:197quotes_checked: long prompt text inside theReplyfield metadata mixes the prompt with the schema
Maintainability senior, hardest first:
- Locality, Shape:
bin/chunk_pr/main.py:298merged: a while loop holding a queue turn, approval waits and adoneset only on some paths, with no loop invariant stated - Locality, Self-sufficiency:
bin/loop_claude/routing.py:113unfinished_route: five rungs, the escalation count and the mark from another file, and each step's reason in the decision log - Locality, Self-sufficiency:
skills/larv-review/SKILL.md:146"Comprehension check": paragraphs of four or five conditions, each pointing to a hook, a CLI's help or another rule file - Shape:
bin/release_plan/release_standing.py:162next_release: three sources of a cut, theis_earliestexpression andLastCutfields typedstr | None | Failure - Shape:
bin/release_plan/ranking.py:136weigh: two live table shapes double every path, andShapeis built from positional tuples of callables - Locality, Shape:
bin/loop_claude/cycle/run.py:151rule: judges, stamps, archives and steps up, beside a 17-fieldCyclepassed around whole - Shape:
bin/chunk_pr/main.py:438remeasure: control flow inside chainedisinstance(x, Failure) or …expressions, andfinishedunpacks five runs by position - Shape:
bin/chunk_preflight/main.py:128preflight: the whole fallback logic packed into lines 142-143
Inherited architect, would restructure first:
- Navigation, Shape:
README.md:193thebin/roles: about 70 flat entries underbin/, where only this list tells a tool from a library - Navigation, Shape:
bin/larv_review/main.py:222COMMANDS: the package named for the review mostly holds the comprehension check, and the review loop is prose in SKILL.md - Locality, Shape:
bin/chunk_pr/main.py:384levelled_and_finished: merges, finishes items, pushes, asks whether a scoring run is owed and remeasures under a name that gives two of the five - Locality:
bin/chunk_pr/main.py:349unplanned: matches the error text oftodo-item show, asfinisheddoes, whereplan_storereads the plan - Navigation:
bin/rule_hook/round_cap.py:10round_cap_breaches: the round cap is enforced in the hook, not inlarv_review, where a reader looks first - Shape:
bin/deploy_release/forge.py:1module: the forge lives indeploy_releaseand inchunk_preflight.choices.forge_of, neither a home its name gives - Shape:
bin/rule_hook/rules.py:26RULES: the file is the registry and also four rules' checkers, so where a new rule goes is undecided - Self-sufficiency:
bin/plan_store/stores.py:5module docstring: cites plan items 365, 366 and 368 as part of the contract, as routing and ladder code cite decisions
2026-11-16T12:09:27Z, PR #560 at 55235a0
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6.5 | 6 | 6 |
| Mid A | 7.5 | 6.5 | 7 | 6.5 | 7 |
| Maintainability senior | 7 | 6 | 6.5 | 6 | 6.5 |
| Inherited architect | 7 | 6.5 | 6 | 6 | 6.5 |
| Mean | 7.13 | 6.13 | 6.50 | 6.13 | 6.50 |
Junior A, hardest first:
- Locality:
skills/larv-review/SKILL.md:90"Re-entry": six rules held at once with the round cap, the release-review caps and the mechanical-answers exception, as dense prose rather than a decision table - Locality, Self-sufficiency:
bin/release_plan/release_standing.py:176next_release: three sources of the cut version and release names replayed from todo.md's history;is_earliestcarries the meaning of "the plan is current" without naming it - Locality:
bin/loop_claude/routing.py:116unfinished_route: five rungs, a counter and an escalation mark from another module; the docstring table makes it tractable - Shape:
bin/release_plan/ranking.py:239score: the OLD_SHAPE and SHAPE split runs through every layer, andweight_ofpacks a three-way rule into one line - Self-sufficiency:
bin/loop_claude/cycle/judge.py:197quotes_checked: six quote kinds each with its own fallback;INTERRUPT_RULESis behaviour the code cannot verify - Locality:
bin/loop_claude/cycle/run.py:154rule: the order of stamping, archiving and stepping matters across three functions - Shape:
bin/loop_claude/cycle/prompts.py:66review_launches: a prompt assembled from five helpers that may each return nothing
Mid A, hardest first:
- Locality:
bin/loop_claude/routing.py:116unfinished_route: five interacting rungs, a counter and an escalation mark; the module table resolves it, the escalation-mark rule must stay open beside it - Locality:
bin/release_plan/release_standing.py:176next_release: four cut cases plus renamed releases; the order of the checks carries meaning the docstrings don't explain - Locality:
bin/chunk_pr/main.py:298merged: a loop around a flock with approval and comment checks and two exception exits;doneassigned only under branches - Self-sufficiency:
bin/loop_claude/cycle/judge.py:197quotes_checked: the prompt text atjudge.py:22is behaviour written as English - Shape:
bin/release_plan/ranking.py:239score: two table shapes double the branches, andweight_of's defect rule needs a second read - Self-sufficiency:
bin/rule_hook/repeat_read.py:114shown_lines: ISO-string time comparisons and harness behaviour nothing else lets a reader check - Locality:
skills/larv-review/SKILL.md:82"Round cap": Done, Cap and Re-entry cross-reference each other and point out to three other files - Shape:
bin/chunk_pr/main.py:407finished: five subprocess results destructured by position
Maintainability senior, hardest first:
- Locality:
bin/release_plan/release_standing.py:176next_release: four regimes at once;carriedreapplied in three places, one relying onzipkeeping order - Locality:
bin/release_plan/ranking.py:239score: one Score depends on five inputs held in other files, and the two shapes fork every function - Locality:
bin/loop_claude/routing.py:116unfinished_route: rungs, a counter and a mark interact; the table resolves most of it - Locality:
bin/loop_claude/loop/chunk_windows.py:107start_chunks: about 40 lines under one flock that check the owner, split running from stopped, compute room, read the plan and filter, with local state tracked by hand - Self-sufficiency:
bin/loop_claude/cycle/judge.py:197quotes_checked: behaviour split between English rules for a model and verbatim checks - Shape:
bin/plan_store/jira_epic.py:356revision_of: indexes an untyped tuple fromscoresatjira_epic.py:530 - Locality:
skills/larv-review/SKILL.md:82"Round cap": dense rules with exceptions pointing to other sections and files - Self-sufficiency:
bin/larv_review/main.py:1the module docstring: the whole contract of ten subcommands in one block
Inherited architect, would restructure first:
- Navigation, Shape:
bin/release_plan/todo_table.py:1the plan model: lives underrelease_plan/, while the plan is also inplan_store/,todo_item/,rank_todo.pyandlarv_review/plan.py, under three names - Locality:
bin/review_rules/round_titles.py:63past_cap: the round cap is split across this,rule_hook/round_cap.py:10,larv-review round-countand SKILL.md's "Round cap" - Shape:
bin/release_plan/named_code.py:5named_paths: nothing in the module is code - Shape:
bin/loop_claude/verdicts.py:11TakenIn.highest_likelihood: L is a likelihood in the field and a level in the prose - Shape:
bin/loop_claude/loop/routing_check.py:20the routing check: "routing" names four concepts acrossmodel_routing/,routing_figures/and this module - Navigation:
README.md:197thebin/package list: the 72-entrybin/level has only the README's two lists as its index - Self-sufficiency:
bin/review_rules/round_titles.py:7CAP: the why is deferred to a decision-log title in a 200 KB file
2026-11-16T12:13:26Z, PR #560 rerun at 55235a0
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7.5 | 6 | 7 | 6.5 | 6.5 |
| Mid A | 7 | 5.5 | 6 | 5.5 | 6 |
| Maintainability senior | 8 | 6.5 | 7 | 7 | 7 |
| Inherited architect | 7 | 6.5 | 6 | 7 | 6.5 |
| Mean | 7.38 | 6.13 | 6.50 | 6.50 | 6.50 |
Junior A, hardest first:
- Locality, Self-sufficiency:
bin/release_plan/release_standing.py:176next_release: fiveLastCutfields, a sentinel, renames and carried items at once; theis_earliestexpression stayed opaque, and the full case table lives only in chunk-preflight's help - Locality:
bin/loop_claude/routing.py:116unfinished_route: the escalation mark spansescalation_mark.py,judge.py,verdicts.py,route.pyandreload.py - Self-sufficiency:
bin/rule_hook/folder_search.py:89skips: mirrors rg's and ag's flag handling, which nothing in the repo lets a reader check - Shape:
bin/release_plan/ranking.py:239score: the old-shape branches run through every function of ranking - Locality:
bin/loop_claude/cycle/judge.py:197quotes_checked: the verdict changes shape in three places after the checks - Navigation:
bin/larv_review/main.py:1the help docstring: 155 lines of rules, with no pointer to the module that enforces each - Locality:
bin/rule_hook/reviewer_rung.py:19SEAT: a launch counts as a review by matching prompt sentences kept in Markdown - Locality:
bin/loop_claude/loop/reload.py:97exec_if_edited: a running loop re-executes itself on any edit and hands its state over as JSON
Mid A, hardest first:
- Locality:
bin/chunk_pr/main.py:298merged:donebound only under some branches, and the approval wait is a fall-through rather than a named branch;levelled_and_finishedreturns a finish failure only after the push - Locality:
bin/loop_claude/routing.py:116unfinished_route: five rungs, a counter and a mark; only the module table makes the step line readable - Self-sufficiency:
bin/release_plan/release_standing.py:176next_release: three kinds of "none", and the reasoncarriedrewrites releases is a dated call with no rule in the code - Shape:
bin/release_plan/ranking.py:239score: two table shapes through every function, andMAINTAIN_WEIGHT's reason in the decision log - Self-sufficiency:
bin/loop_claude/cycle/judge.py:197quotes_checked: the judge's prompt is spread over field metadata, rule constants and a sentence fromescalation_mark.py - Locality:
skills/larv-review/SKILL.md:35"§3. Passes": control flow as prose whose sub-sections point at each other, enforced only partly by code - Locality:
bin/rule_hook/rules.py:26RULES: each rule's spec is a paragraph apart from the module that implements it - Self-sufficiency:
bin/larv_review/main.py:1the module docstring: the only full spec of a dozen subcommands
Maintainability senior, hardest first:
- Locality:
bin/release_plan/release_standing.py:176next_release: state from git history, a sentinel, andcarriedrun twice, whose agreement must be checked by hand - Locality:
bin/loop_claude/routing.py:116unfinished_route: five rungs, a count and a mark; table cells need thestepsdocstring to parse - Shape:
bin/release_plan/ranking.py:144weigh: two table shapes through every step, andScoring.weightsmeans different things per shape - Shape:
bin/plan_store/jira_epic.py:356revision_of: indexes the untyped tuplescoresreturns by position - Locality:
bin/loop_claude/loop/chunk_windows.py:107start_chunks: a flock, running and dead windows,leaving,room, the restart of stopped windows, held items and a NoItem fallback in one function, with the limit applied again inmay_start - Locality:
bin/chunk_pr/main.py:298merged: whetherdoneis bound when read depends on threeif readybranches lining up - Self-sufficiency:
bin/loop_claude/cycle/judge.py:22INTERRUPT_RULES: what the model is asked and what code checks need both files open - Self-sufficiency:
bin/rule_hook/rules.py:44RULES: the env-file-read spec is a 120-word string kept in step with code elsewhere
Inherited architect, would restructure first:
- Locality, Shape:
bin/plan_store/stores.py:1the plan store:release_planandplan_storeimport each other, and the goals file sits underrelease_plan - Shape:
bin/larv_review/plan.py:22scoring_run_due: the scoring-run trigger is spread overplan.py,owed.py,unmeasured.pyandscoring.py, andplan.pyisn't about the plan - Locality:
bin/review_rules/round_titles.py:45the round cap: constants, commit enforcement inrule_hook/round_cap.py:10, the launch refusal inlarv_review/reviewer_header.py:73, the printout and SKILL.md's prose - Shape:
bin/loop_claude/cycle/run.py:55Cycle.todo: says todo.md where the plan may be a Jira epic, asfetched_todoandINTERRUPT_RULESdo - Shape:
bin/loop_claude/hand_over.py:23hand_over: shares its name with the reload's handover of state inloop/reload.py:36decode_handed - Locality:
bin/loop_claude/loop/close.py:18closed_cycle: a second "close" besidecycle/run.py:234close - Navigation, Shape:
README.md:197thebin/package list: 74 entries at one level with the README's list as the only index - Self-sufficiency:
bin/loop_claude/loop/routing_check.py:1the module docstring: namesstate.CYCLES_PER_CHECKwithout saying which modulestateis
2026-11-16T15:57:39Z, PR #562 at 31da587, against 64454f8
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | better | better |
Mid A, decided by:
- Locality, Shape:
bin/plan_store/stores.py:1store: the 234-line module holding the protocol, the markdown backend, the Jira scoring and a deferred import of jira_epic is now a 40-line factory besideplans.py,todo_md.pyandscoring.py, with no cycle to keep in mind - Locality:
bin/record_store/stores.py:34RecordStore.write: the Confluence store's hiddenseenandkeptfile gave way toheld()andwrite(text, over), so a save is followed without knowing the store's history - Navigation, Shape:
bin/release_plan/next_chunk.py:57planned: the next chunk's pick moved out of chunk-preflight's internals beside the release standing, and itsPlanbecameNextChunk, no longer clashing with the store'sPlan - Locality, Shape:
bin/release_plan/release_standing.py:169standing_of: the plan is weighed against the cut once, wherecarried,cut_versionandnamed_releasesran in three places - Navigation, Shape:
bin/atlassian/client.py:1client: Confluence's auth no longer lives in the Jira plan backend - Navigation:
bin/notifier/notify.py:10BACKENDS: both DM backends sit in the notifier's folder - Shape:
bin/rank_todo.py:119main: one write path, throughFilePlan.rank(), in place ofplan_file.py's copy
Inherited architect, decided by:
- Locality, Shape:
bin/record_store/confluence_page.py:84ConfluencePage.write: the concurrency check shows in the call signature, where it hung on which earlier read ran in which process - Shape, Locality:
bin/repo_checks/test_package_cycles.py:13package_imports: the deferred imports that hid the plan packages' cycles are gone, a neutralatlassianpackage holds the shared client, and a test enforces the rule - Navigation, Shape:
bin/release_plan/next_chunk.py:57planned: the pick no longer lives in one CLI's package, and its type is namedNextChunk - Shape:
bin/plan_store/stores.py:18store: the store is split by purpose, andrelease_plan/is no longer a grab bag - Locality:
bin/release_plan/release_standing.py:169standing_of: weighed once, withplan_is_currentandpick_fromnaming what they do - Locality:
bin/plan/todo_table.py:294item_fields: one check of an item's cells for every backend - Locality:
bin/plan_store/plans.py:57no_item: the coupling between chunk-pr and todo-item's message has a name - Locality:
bin/chunk_pr/main.py:130step: the config is passed in once, wheremisfitshelled out toproject-configtwice - Shape:
bin/project_config/settings.py:320SCALARS: one row per one-value setting, whereloadhandled each at four sites - Navigation, Self-sufficiency:
bin/chunk_preflight/gathered.py:60Gathered.plan: a stale pointer tochoices.plan_name, with two more innext_chunk.pyand the README, small next to the gains
2026-11-19T09:23:24Z, PR #633 at 81e8d8a1
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 6 | 6 | 5.5 | 6 |
| Mid A | 7 | 6 | 6.5 | 5.5 | 6.5 |
| Maintainability senior | 7.5 | 6.5 | 6.5 | 6.5 | 7 |
| Inherited architect | 7 | 7 | 6 | 6 | 6.5 |
| Mean | 7.13 | 6.38 | 6.25 | 5.88 | 6.50 |
Junior A, hardest first:
- Locality:
bin/do_epic/main.py:283parked_through: one polling loop carries five mutable locals against anowof four types, and a nested conditional must be traced by hand - Locality:
bin/do_epic/scheduler.py:136start_chunks: four overlapping sets feed an elif chain whose arm order decides the outcome, under a lock with a second one inside - Self-sufficiency:
skills/larv-review/SKILL.md:97"Re-entry": dense rules held at once with the round cap, pointing out to other files - Locality, Self-sufficiency:
bin/agent_cycle/routing.py:115unfinished_route: five rungs, a counter and a mark live at once; the why of the caps is only in the decision log - Self-sufficiency:
bin/agent_cycle/judge.py:235quotes_checked: six quote checks each with its own fallback; the prompt constants are behaviour guarded by one comment - Locality:
bin/loop_claude/loop/main.py:108next_cycle: the order of steps carries meaning the code doesn't state - Locality:
bin/chunk_pr/main.py:429levelled_and_finished: order-dependent steps that comments state and nothing enforces - Self-sufficiency:
bin/rule_hook/rules.py:27RULES: a rule's only specification is one long sentence with nested unless clauses
Mid A, hardest first:
- Locality:
bin/agent_cycle/routing.py:115unfinished_route: interacting rungs, a counter and an escalation mark; the module table resolves it - Self-sufficiency:
bin/agent_cycle/judge.py:235quotes_checked: the prompt text is behaviour written as English - Shape:
bin/loop_claude/cycle/prompts.py:132escalation: a prompt assembled from pieces that may each be empty - Locality:
bin/agent_cycle/run.py:150rule: the order of stamping, archiving and stepping matters across functions - Self-sufficiency:
bin/rule_hook/rules.py:27RULES: rule text doubles as specification and help - Shape:
bin/rule_hook/endless_wait.py:93self_match: long lines and a regex whose cases need a second read - Locality:
bin/loop_claude/loop/chunk_windows.py:107may_start: depends on state kept out of sight - Self-sufficiency:
bin/larv_review/plan.py:80plan_reasons: reasons whose meaning lives in the skill's text - Locality:
bin/model_routing/route.py:42Route: fields used as archive keys elsewhere; the unit this seat least wanted to modify
Maintainability senior, hardest first:
- Locality:
bin/do_epic/main.py:283parked_through: five mutable locals in one polling loop, and an empty string that means "never parked" - Locality:
bin/chunk_pr/main.py:429levelled_and_finished: order-dependent steps, and a context manager that yields a Failure instead of raising - Locality, Shape:
bin/do_epic/scheduler.py:136start_chunks: overlapping sets and an elif chain under nested locks - Self-sufficiency:
bin/agent_cycle/routing.py:115unfinished_route: the why of the caps sits in the decision log - Self-sufficiency:
bin/agent_cycle/judge.py:235quotes_checked: fallbacks change the verdict's type without a signal at the call site - Self-sufficiency:
bin/rule_hook/rules.py:27RULESenv-file-read: a 130-word sentence is the rule's only specification - Shape:
bin/plan/ranking.py:239score: two table shapes double the branches - Locality:
bin/agent_cycle/run.py:240close: closing steps whose order matters across the module
Inherited architect, would restructure first:
- Shape:
bin/root: about 85 entries at once with no split between tools and libraries - Locality:
bin/do_epic/main.py:11imports: do_epic reaches into loop_claude's internals - Navigation:
bin/loop_claude/cycle/the folder: the name misleads after its code moved to agent_cycle - Navigation:
bin/agent_cycle/verdicts.py:1the module: names left from before the move - Shape:
bin/agent_cycle/run.py:38Cycle: the config, judge and project fields widen a 16-field value - Shape:
bin/larv_review/reviewer_header.py:121Launch: built positionally - Self-sufficiency:
bin/larv_review/plan.py:25the module: no docstring says what it holds
2026-11-19T09:27:43Z, PR #633 rerun at 81e8d8a1
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 6.5 | 5 | 6 | 4.5 | 5.5 |
| Mid A | 7 | 6 | 6.5 | 6 | 6.5 |
| Maintainability senior | 7 | 5.5 | 6 | 5.5 | 6 |
| Inherited architect | 6.5 | 6 | 5 | 5 | 5.5 |
| Mean | 6.75 | 5.63 | 5.88 | 5.25 | 5.88 |
Junior A, hardest first:
- Locality, Self-sufficiency:
bin/agent_cycle/routing.py:115unfinished_route: five rungs, a count and the escalation mark held at once; the why sits in the decision log andescalation_mark.pycheckedmust be open beside it - Self-sufficiency:
bin/agent_cycle/judge.py:235quotes_checked: six kinds of quote each fail differently, and half the behaviour is prompt text whose load-bearing sentences can't be told apart - Self-sufficiency:
skills/larv-review/SKILL.md:97"Re-entry": rules with exceptions to exceptions, cross-referencing other sections and files - Locality, Shape:
bin/agent_cycle/run.py:150rule: a 16-fieldCycle, long one-line calls, and an archive-before-step-up order only reading order tells - Self-sufficiency:
bin/rule_hook/rules.py:27RULES: each rule is one enormous sentence, and its check lives in another module - Locality:
bin/loop_claude/loop/chunk_windows.py:107may_start: sets that depend on tmux and lock state out of sight, a lock a second program also takes - Locality:
bin/loop_claude/loop/reload.py:69decoded_route: renaming a field stops running loops, and nothing warns at the other end - Locality:
bin/do_epic/main.py:118run: argv parsed through interacting booleans, and imports of loop_claude internals
Mid A, hardest first:
- Locality:
bin/do_epic/main.py:283parked_through: a polling loop holds four variables across three result types, a three-deep chained conditional, andon is Nonedoubling as the first pass; the unit this seat least wanted to modify - Locality, Self-sufficiency:
bin/agent_cycle/routing.py:115unfinished_route: five rungs plus a count and a mark held together; the why is cited to the decision log - Locality:
bin/do_epic/scheduler.py:136start_chunks: four window sets feed a five-arm chain whose meaning changes withown, inside two nested flocks shared with loop-claude - Self-sufficiency:
bin/agent_cycle/judge.py:235quotes_checked: six quote kinds fail differently; much behaviour is an English prompt, andTASK_HEADINGis silently an archive format - Locality:
bin/project_config/settings.py:182load: a 20-entry list of cross-setting rules, and an assert that depends on it being complete - Locality:
bin/chunk_pr/main.py:429levelled_and_finished: a load-bearing order and a short-circuit hidden infailed or run(...) - Self-sufficiency:
bin/rule_hook/rules.py:27RULES: each contract is a long English sentence, and nothing ties it to the checking code in eight sibling modules - Shape:
bin/agent_cycle/stuck_workers.py:52Watch.check: the condition at line 59 mixesorandandwithout parentheses over dicts mutated across ticks
Maintainability senior, hardest first:
- Locality:
bin/do_epic/main.py:283parked_through: four mutable locals across iterations, a side effect correct only becauseonis reassigned later, and a crash-ordering invariant across three stores; the unit this seat least wanted to modify - Locality:
bin/do_epic/scheduler.py:136start_chunks: four overlapping sets from two stores feed a five-arm ladder, and a parked chunk silently restarts down the no-repo path - Locality:
bin/loop_claude/loop/chunk_windows.py:107may_start: depends on window names and options, dead flags, archived lines, a flock and predictions, with the schema inagent_cycle/tmux.py - Locality:
bin/agent_cycle/routing.py:115unfinished_route: five rungs plus a count and the mark; the table resolves it, the why is in the decision log - Locality:
bin/chunk_pr/main.py:429levelled_and_finished: a load-bearing order held together by one-liners where a Failure's truthiness carries the control flow - Shape:
bin/agent_cycle/run.py:150rule: archive-then-step-up order, several exits, and very long single-line expressions - Locality:
bin/project_config/settings.py:182load: one list encodes the whole validation matrix, whose interactions don't read - Self-sufficiency:
bin/agent_cycle/judge.py:235quotes_checked: six quote kinds with different fallbacks;INTERRUPT_RULESis behaviour no code can check
Inherited architect, would restructure first:
- Locality, Self-sufficiency:
skills/larv-review/SKILL.md:89"Round cap": the cap is spread across three packages and this prose, and nothing in code counts a reviewer's launches - Locality:
bin/do_epic/main.py:11imports: do_epic reaches into eight loop_claude modules, over a wire format a renamed field breaks - Navigation:
bin/loop_claude/cycle/the folder: it holds prompts and chunk items while the cycle is agent_cycle - Shape:
bin/agent_cycle/verdicts.py:1the module: its docstring names acycle.package that is now agent_cycle - Shape:
bin/root: about 80 entries with no grouping, indexed by one dense README paragraph - Navigation:
bin/loop_state/paths.py:6Paths: neither state nor loop-specific - Self-sufficiency:
bin/larv_review/plan.py:25scoring_run_due: the module holds the scoring-run trigger and has no docstring - Locality:
bin/review_rules/review_passes.py:8PASS_AGENTS: code constants that must match sentences in SKILL.md
2026-11-19T13:08:12Z, PR #634 at 914b8dc2, against aedc5ed3
Ruling: better
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | better | better | better | same | better |
| Inherited architect | better | better | better | same | better |
Mid A, decided by:
- Navigation, Shape, Overall:
bin/do_epic/main.py:11next_cycle: do-epic imports the sharedchunk_driverinstead of loop-claude's internals, so a do-epic routing symptom no longer leads into loop-claude'smain.py - Shape, Locality:
bin/chunk_driver/chunk_run.py:44closed_cycle: takes the task as an argument, and do-epic runs onChunkRuninstead of aLoopRunfilled with empty fields - Navigation, Shape:
bin/chunk_driver/cycle_start.py:96cycle_of: both entry points build a cycle through the same three steps, which the module docstring lists in order - Navigation:
bin/forge/reviews.py:23review_decision: one module reads a pull request's review state, where two tools each read part of it - Navigation:
bin/atlassian/jira.py:1module docstring: Jira search and transitions sit in the Atlassian package, not in the plan store - Locality:
bin/wait_for/main.py:30main: the clock is a visible parameter instead of a patched global
Inherited architect, decided by:
- Navigation, Locality, Shape, Overall:
bin/do_epic/main.py:410next_cycle: calls the sharedproject_routing,announced_routeandcycle_of, so how a cycle starts changes in one file, not in two kept in step - Locality:
bin/do_epic/main.py:439chunk_work: returnsEpicReopenedinstead of handing chunks on from inside a getter - Shape, Locality:
bin/chunk_driver/chunk_run.py:33ChunkRun: holds only what do-epic needs; theRunprotocol's use of__dataclass_fields__is clever but commented - Navigation, Shape:
bin/chunk_driver/claims.py:1module docstring: the lock and the chunk flags sit in one module named for that job, andstartingyields aFailureinstead of raising - Locality, Navigation:
bin/forge/reviews.py:23review_decision: all review-state reading inforge/, each field shape-checked; three similar names cost a docstring read each - Locality:
bin/chunk_driver/project/model_choice.py:28overridden: the model override is one function, where it was spelled out twice
2026-11-19T22:33:15Z, PR #646 at a357cc7b
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 7 | 5.5 | 6 | 6 | 6 |
| Mid A | 7 | 6 | 6.5 | 6 | 6.5 |
| Maintainability senior | 7 | 6 | 6 | 6 | 6.5 |
| Inherited architect | 7 | 6.5 | 6 | 6 | 6.5 |
| Mean | 7.00 | 6.00 | 6.13 | 6.00 | 6.38 |
Junior A, hardest first:
- Locality, Shape:
bin/do_epic/main.py:287parked_through: the three-step park and resume order, thetoldandfinishedflags and a raw string compare held at once; the docstring gives the order but not why the flags exist - Locality, Self-sufficiency:
bin/agent_cycle/routing.py:115unfinished_route: floor rung, escalation mark, cap and the count of unfinished verdicts meet in a dozen lines, and the reasons sit in the decision log - Shape:
bin/agent_cycle/judge.py:237quotes_checked: six checks each quietly turn one verdict type into another, and the call site doesn't show the result type - Locality:
bin/do_epic/scheduler.py:144start_chunks: four overlapping sets, a five-branch chain, a counter and nested locks shared with loop-claude - Shape:
bin/agent_cycle/run.py:150rule: runs the judge, stamps, annotates, steps up and orders the archive writes, beside four near-identical archive helpers - Locality:
bin/agent_cycle/stuck_workers.py:52Watch.check: mutatesself.cpuandself.judgedout of sight, and one condition mixesandandorwithout parentheses - Shape:
bin/do_epic/gates.py:42gate: a long ordered chain where a missing value means something different on each line - Locality:
bin/loop_claude/loop/reload.py:36decode_handed: pops legacy fields from a handover the old process wrote before it replaced itself; nothing enforces the coupling - Self-sufficiency:
skills/larv-review/SKILL.md:89"Round cap": three sections must be open at once to tell whether a fix reruns a pass - Self-sufficiency:
bin/agent_cycle/verdicts.py:41the comment: points tocycle.judge.re_ask_checked, a module path that doesn't exist
Mid A, hardest first:
- Locality:
bin/do_epic/main.py:354resumed_news: awhile Truepoll that carriessaidandshownacross turns, with a nested-quote f-string; no comment says what holds when the loop exits - Locality, Self-sufficiency:
bin/do_epic/main.py:287parked_through: three steps whose correctness depends on each other; the reason for the order sits in the decision log - Locality:
bin/do_epic/main.py:241chunk_run:news,told,finishedand the reopened-epic exit held across iterations - Locality:
bin/do_epic/scheduler.py:144start_chunks: a five-branch chain over four sets, counted under nested locks - Shape:
bin/project_config/settings.py:187load: about twenty comprehension clauses build onefaultslist, so finding what one setting requires means scanning all of them - Locality:
bin/agent_cycle/routing.py:115unfinished_route: five rung concepts at once, readable only withbin/model_routing/ladder.pyopen beside it - Shape:
bin/chunk_pr/main.py:447levelled_and_finished: checks whether a scoring run is owed, which its name doesn't suggest - Locality:
bin/plan/ranking.py:239score: the formula forks on the old and the new table shape, both live - Self-sufficiency:
bin/larv_review/main.py:1the module docstring: about 165 lines that restate whatCOMMANDSand the modules it calls enforce
Maintainability senior, hardest first:
- Locality, Shape:
bin/do_epic/main.py:287parked_through: the told, finished and news interplay, the human-merge cases and the three-step order held at once, and no code enforces the order - Locality, Shape:
bin/do_epic/main.py:354resumed_news: a loop that mutates five locals and builds its message in a nested f-string, with no stated invariant - Locality:
bin/agent_cycle/routing.py:115unfinished_route: five rungs, an unfinished count, the escalation mark and the cap held together; each threshold's reason sits in the decision log - Shape, Self-sufficiency:
bin/project_config/settings.py:187load: about 20 comprehension expressions each encode a rule linking two settings; named predicates would make each rule findable - Shape, Locality:
bin/chunk_pr/main.py:314merged: the merge-queue loop with its flags, where a variable carries either a string or a Failure and each read must work out which - Locality:
bin/plan/ranking.py:239score: the old and new table shapes run through the whole module, so each change is checked against both - Locality:
bin/larv_review/plan.py:21Emptied: astrsubclass whose type carries meaning, lost without warning by any code that rebuilds the string - Locality:
bin/agent_cycle/judge.py:76TASK_HEADING: an archive key, so rewording it breaks replays in another module, and only a comment warns - Self-sufficiency:
skills/larv-review/SKILL.md:89"Round cap": the review rules exist three times, here, in the help atbin/larv_review/main.py:1and in rule-hook's code
Inherited architect, would restructure first:
- Locality, Shape:
bin/rule_hook/round_cap.py:10round_cap_breaches: the round cap lives in four places across three packages, and none owns the whole invariant; rounds are counted from commit titles, never from launches - Navigation, Shape:
README.md:225"Packages under bin/": 84 entries listed alphabetically in dense prose, with no grouping by layer - Shape:
bin/loop_claude/loop/reload.py:18LoopRun: loop-claude's core run state defined in a module named for self-reloading - Shape:
bin/agent_cycle/judge.py:116Rejected: says loop-claude rejected the reply, in a module do-epic shares - Self-sufficiency:
bin/agent_cycle/run.py:103the comment: citesclose.py, which doesn't exist - Self-sufficiency:
bin/agent_cycle/verdicts.py:41the comment: citescycle.judge.re_ask_checked, a stale module path - Navigation:
bin/chunk_driver/drivers.py:1the module: "entry point" and "driver" name one concept, andRun,ChunkRunandLoopRunare one concept in three types - Navigation:
bin/do_epic/status.py:1the module: in no map, and the package's__init__.pyis empty like 40 others - Locality:
reviewer-merge-trials/driver.py:4the imports: a frozen trial imports live code underbin/, so a refactor there breaks a record - Self-sufficiency:
bin/agent_cycle/routing.py:92the comment: one of 109 citations of the decision log that carry a unit's reason
2026-11-19T23:28:20Z, PR #645 at 2797e110, against a9bdef6
Ruling: same
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Mid A | same | worse | same | same | same |
| Inherited architect | same | worse | same | same | same |
Mid A, decided by:
- Locality:
bin/plan/todo_table.py:110Item.awaits_merge: the plan core's item carries a mark that release_plan.awaiting_merge sets from forge state, so whether an item waits depends on whether project_plan's caller passed a forge_run. - Locality:
bin/release_plan/next_chunk.py:88planned: the loop's stop is the truthiness of a sentence, and travels through chunks_to_start, AwaitedMerges, start_chunks, windows_opened and handed_on. - Locality:
bin/do_epic/main.py:431handed_on: one straight path became two booleans that three call sites set, branching three ways against config.automation_moves.
Inherited architect, decided by:
- Locality:
bin/release_plan/release_standing.py:104project_plan: a one-line delegate that read git now calls the forge behind a four-part guard, while the module docstring speaks only of the default branch. - Locality:
bin/plan/todo_table.py:110Item.awaits_merge: a field nothing in its own package sets; waiting reads it, so it depends on a caller three packages out having run the marking. - Locality:
bin/loop_claude/loop/chunk_windows.py:161chunks_to_start: the stop condition is split across two packages, and one half is a string's truthiness. - Locality:
bin/do_epic/scheduler.py:96open_blockers: four functions gained a defaultedunblocking, so a caller that forgets it silently gets the old behaviour.
2026-11-22T15:29:07Z, item-503 at 16d9b3a0
| Seat | Navigation | Locality | Shape | Self-sufficiency | Overall |
|---|---|---|---|---|---|
| Junior A | 6 | 5 | 5.5 | 5 | 5.5 |
| Mid A | 7 | 5.5 | 6 | 5 | 6 |
| Maintainability senior | 7 | 5.5 | 6 | 6 | 6 |
| Inherited architect | 7 | 6.5 | 6 | 7.5 | 7 |
| Mean | 6.75 | 5.63 | 5.88 | 5.88 | 6.13 |
Junior A, hardest first:
- Navigation, Overall:
bin/agent_cycle/run.py:102archived: cites a close.py that doesn't exist, which cost a search - Locality, Shape, Overall:
bin/do_epic/main.py:538handed_on: five-way branch over outcome, automation and merged-status config - Locality:
bin/do_epic/scheduler.py:172start_chunks: chunk state is rebuilt from tmux window names no module owns - Self-sufficiency:
bin/agent_cycle/run.py:32Cycle: the why of its 17 fields sits in docs/decisions.md
Mid A, hardest first:
- Locality, Overall:
bin/chunk_pr/main.py:587levelled_and_finished: step order is load-bearing and unmarked - Shape, Overall:
bin/do_epic/main.py:539handed_on: five branches over three configuration facts - Self-sufficiency:
bin/larv_review/main.py:1module docstring: the rules for votes and owed runs live only in a 167-line help text
Maintainability senior, hardest first:
- Locality, Overall:
bin/do_epic/main.py:379parked_through: starts other chunks from what reads like a wait - Locality:
bin/agent_cycle/tmux.py:92as_listed: chunk state shared across processes in window names - Shape:
bin/plan/ranking.py:239score: the OLD_SHAPE fork doubles the paths through ranking
Inherited architect, would restructure first:
- Navigation, Shape, Overall:
README.md:234"Packages under bin/": about 85 entries at one level, six packages missing from the list - Locality:
bin/rule_hook/main.py:94breaches_of: the round cap binds at commit time, not at reviewer launch - Shape:
bin/chunk_pr/main.py:167step: seven subcommands and inline forge branches in one 764-line module