State what a pass's round is and that the round cap counts every one
This commit was merged in pull request #111.
This commit is contained in:
@@ -17,7 +17,7 @@ Decisions live in `docs/decisions.md`:
|
||||
- loop-claude: a CR-only README.md is reported at line 1
|
||||
- loop-claude: each project's model lives in loop-claude's own file
|
||||
- rebuild: loop-claude and the review process are rebuilt on `rebuild` and merge to main whole
|
||||
- review: the round cap counts a pass's runs per chunk
|
||||
- review: the round cap counts a pass's rounds per chunk
|
||||
- loop-claude: the judge is retried alone, on its own counter
|
||||
- loop-claude: the judge loads `CLAUDE.md`
|
||||
- loop-claude: the archive keeps the raw transcript, and a failed archive stops the loop
|
||||
|
||||
@@ -13,7 +13,7 @@ Shorter is better, none is best — chat output, commit messages, comments, docs
|
||||
- **premises** — the goals and audience under `## Goals` and `## Audience` in the project's `README.md`.
|
||||
- **release** — a semver version the human cuts.
|
||||
- **reviewer** — a reporter of findings: an agent under `~/.claude/agents/`, `/security-review` or a forge's auto-reviewer.
|
||||
- **round cap** — the runs one pass may take per chunk, re-entries included.
|
||||
- **round cap** — the rounds one pass may take per chunk.
|
||||
- **session** — a Claude Code session: one id, what `claude --resume` continues.
|
||||
- **trigger** — the condition under which a review pass runs on a change it has not seen; `larv-review` §3 lists them.
|
||||
- **worker** — the headless claude that runs the task in a cycle, in the chunk's session.
|
||||
|
||||
+2
-2
@@ -52,9 +52,9 @@
|
||||
|
||||
2026-09-28, lilleman. Serves goal 4. Three scoring runs held Locality at 6 while about thirty chunks fixed the places the seats named, and two blind judges scored main's loop-claude 5 and its review process 4 against 7 for sketches designed from the target result. Other sessions run on `~/.claude` main throughout, so every rebuild chunk merges into `rebuild`, never main, and `rebuild` reaches main only when it passes the comprehension floor and **the human has explicitly approved the whole and merges it themselves**. The ⚠️ warning stands in `AGENTS.md`, atop `todo.md` and on its last item on purpose, 2026-09-28, lilleman: a premature merge lands under every session running on main, so a reader of any one of them must meet it. Valid until the rebuild merges.
|
||||
|
||||
## review: the round cap counts a pass's runs per chunk
|
||||
## review: the round cap counts a pass's rounds per chunk
|
||||
|
||||
2026-09-28, lilleman, for the rebuild. Serves goal 4. A pass runs at most three times per chunk, re-entries included, and a fourth run asks the human; one counter, and no fix bounces between two passes without end. Valid while one human answers the ask.
|
||||
2026-09-28, lilleman, for the rebuild. Serves goal 4. Every round of a pass counts against one cap per chunk, and a fourth asks the human; one counter, and no fix bounces between two passes without end. Valid while one human answers the ask.
|
||||
|
||||
## loop-claude: the judge is retried alone, on its own counter
|
||||
|
||||
|
||||
@@ -30,7 +30,6 @@ A review's first line, derived from the findings:
|
||||
- Surface everything in round 1.
|
||||
- From round 2 the prompt carries the previous round's findings verbatim and what changed since. Verify each earlier 🔴 and 🟡 is resolved — re-run the scenario, never trust the claim — then review what changed. A fix raises a new finding only where the fix itself introduced a real problem.
|
||||
- A pass is done on no 🔴 and no 🟡. The 🟢 and ❓ left are addressed in that round — fixed, answered, declined or filed in `todo.md` — and take no further round, unless a fix changes behaviour: that takes one.
|
||||
- Round cap: a pass runs at most three times per chunk, re-entries included. Instead of a fourth, list what is open and ask the human.
|
||||
|
||||
## Declining
|
||||
|
||||
|
||||
@@ -39,6 +39,8 @@ A comment is prose, not code. Passes run in this order: line-level polish on cod
|
||||
|
||||
**Re-entry.** A fix is a change the passes have not seen. The pass whose findings it answers runs again, per the findings document's rounds; beyond it, classify the fix by the widest thing it touches — the first row whose trigger it fires, none sending it to §4 — and run forward from that row through §4, each later pass only where its trigger fires too.
|
||||
|
||||
**Round cap.** A pass takes at most three rounds per chunk; instead of a fourth, list what is open and ask the human. A round is one launch of a pass's reviewer, whatever launched it: the trigger, a re-entry, a 🔀 escalation, a Stage 0 launch and its rerun. The architecture pass's round is the architect and then the scoring run, so a floor fix reruns both as one round; the prose pass's round is its fresh-context agent. The external auto-reviewer (§4) is a pass of its own, each review it posts a round: after its third, ask the human before any push that would summon a fourth.
|
||||
|
||||
### Architecture
|
||||
|
||||
Launch systems-architect: review the system as a whole entered through this change; the decision log carries previously accepted decisions to honor, the README the goals and Stage 0 artifacts.
|
||||
@@ -84,7 +86,7 @@ The bar is `~/.claude/CLAUDE.md` → "Comments and documentation"; per paragraph
|
||||
2. Watch the forge's checks until every one passes. A red check is a finding; its fix re-enters per §3.
|
||||
3. No CI configured → the gate is local tests only; note it in the summary.
|
||||
4. Everything green → mark the PR ready, dropping any `WIP:`/`Draft:` title prefix.
|
||||
5. **External auto-reviewer**, when the repo has one — its config file (`.coderabbit.yaml` and the like), or its comments on earlier PRs; none → skip it, say so. Marking the PR ready is what summons it. Wait up to **10 minutes** for its first review, checking about once a minute; nothing by then → note it in the summary and carry on, never waiting twice. Its findings follow the findings document like any reviewer's, and its reviews count against the round cap as a pass's runs do. Once it has reviewed, each later return here, after a fix re-entered per §3, waits up to **5 minutes** for its re-review of the new commit.
|
||||
5. **External auto-reviewer**, when the repo has one — its config file (`.coderabbit.yaml` and the like), or its comments on earlier PRs; none → skip it, say so. Marking the PR ready is what summons it. Wait up to **10 minutes** for its first review, checking about once a minute; nothing by then → note it in the summary and carry on, never waiting twice. Its findings follow the findings document like any reviewer's. Once it has reviewed, each later return here, after a fix re-entered per §3, waits up to **5 minutes** for its re-review of the new commit.
|
||||
|
||||
## 5. Goal alignment, then merge
|
||||
|
||||
@@ -95,4 +97,4 @@ The bar is `~/.claude/CLAUDE.md` → "Comments and documentation"; per paragraph
|
||||
|
||||
## 6. Wrap up
|
||||
|
||||
Summarize: PR URL and final state, every reviewer's verdict, runs per pass, passes skipped, findings fixed vs declined vs open — each decline with the goal it rests on — CI status, decisions recorded (file paths), `todo.md`/`CHANGELOG.md` updated, and any 🔀 escalation left for the user.
|
||||
Summarize: PR URL and final state, every reviewer's verdict, rounds per pass, passes skipped, findings fixed vs declined vs open — each decline with the goal it rests on — CI status, decisions recorded (file paths), `todo.md`/`CHANGELOG.md` updated, and any 🔀 escalation left for the user.
|
||||
|
||||
@@ -8,7 +8,6 @@ The rebuild lifts every dimension: each item below is one chunk on a worktree ba
|
||||
|
||||
The places the panel on `rebuild` at 37edc9a named (`docs/decisions.md`), capping dimension first:
|
||||
|
||||
- **State in larv-review's §3 what a pass's run is, and whether a Stage 0 rerun, a panel rerun after a floor fix, an auto-reviewer re-review and a 🔀 escalation each count against the round cap.** Four seats reread the re-entry rule with the table, `principles/review-findings.md` → "Rounds" and the Vocabulary open; round and run name one thing.
|
||||
- **Point `agents/stability-reviewer.md`'s 🔀 list at larv-review's pass table.** Its security trigger (auth/session, payment, file upload) disagrees with the table's (authn/authz, secrets, crypto, untrusted input).
|
||||
- **Return the verdict from `run_cycle`'s judge–interrupt loop, so nothing reads `verdict` after it.** Three seats; `archive`'s skip on an existing copy holds only because each interrupt appends a turn.
|
||||
- **Name the minute `strike` sleeps, and the exec `reload_if_edited` may not return from, where each is called.** Four seats and three.
|
||||
|
||||
Reference in New Issue
Block a user