Release review of 0.11.0 #562

Merged
lilleman merged 15 commits from release-review-0.11.0 into main 2026-11-16 17:01:08 +01:00
Owner

Release review of 0.11.0: the architecture and principles passes over the whole project, and the fixes for the drift they found.

What changed

  • The plan's packages import one way: a pure bin/plan/ core, bin/atlassian/ for the Jira and Confluence client, and plan_store split into plans, todo_md, scoring and stores. A repo check fails on any cycle among bin/'s packages.
  • One parse and one print of an item's cells, item_fields and row_cells, for every plan backend.
  • A pure room_of decides which chunks start; the settings are a table, SCALARS; Standing weighs the plan against the cut once; the next chunk's pick is release_plan/next_chunk.py's NextChunk; loop-claude reads the plan once per cycle start.
  • chunk-pr loads the config in-process, and the title rule has one home, templates.misfits.
  • Record stores keep no hidden version: held() and write(text, over). lilleman chose this on 2026-11-16, and the decision is reworded to match.
  • rule-hook's run-heavy rule also catches runs through uv run, poetry run, sudo and doas. lilleman chose this on 2026-11-16.
  • The notifier backends live in notifier/, and one definition covers the changelog's unreleased heading.
  • lilleman's L 3 / L 4 routing ruling is filed as item 393 for 0.12.0, by their choice. The regrouping of bin/ is filed as item 392 for 0.14.0.

Reviews

  • Architecture: 3 rounds. BLOCK in rounds 1 to 3; lilleman confirmed the restructure, and each round's findings are fixed.
  • Principles: 2 rounds. Its findings are fixed, filed or declined below.
  • Security: /security-review found nothing. The full claude-security scan could not run in a subagent.
  • Overrides: 2 rounds, approved.
  • Stability: 1 round, 14 🟡, no 🔴; all are fixed, three of them as places, below.
  • Product-owner: 1 round, 2 🟡, fixed.
  • Prose: 1 round, 9 🟡, fixed.
  • No auto-reviewer is configured.

Comprehension check: paired ruling better at 31da587 against 64454f8. Mid A voted Navigation, Locality and Shape better, Self-sufficiency same, Overall better. The inherited architect voted better on all four and overall. No scoring run was due.

Overrides: todo-rank-check on the commits that changed todo.md. Reason: the branch's own rank_todo.py needs 0.11.0's required standards setting, which this machine's config can't hold until 0.11.0 is deployed, since v0.10.1's tools refuse it. Both the released and the branch's checker passed. Ruling: approved; the overrides on empty commits are moot.

Places left as they were, and why: loop_claude/cycle/run.py rule, loop_claude/cycle/judge.py INTERRUPT_RULES and larv_review/plan.py scoring_run_due each need a restructure outside this release's themes. Floor item 300 holds them.

🟢 declined

  • OTLP traces: loop-claude's archive is this CLI setup's trace of each cycle. Recording it as a decision would need an item that ends the departure, which is lilleman's call.
  • The two Http aliases are each used only inside their own module.
  • Scalar.read is typed Any, since a mixed table can't tie each reader's type to its field.
  • deploy-release and adf-codec print progress from their imperative shells.
  • chunk-pr's size is item 287's to cut.
  • adf_to_markdown is the codec's one-document API.
  • Item 300's quoted places are left for its next refresh.

Upgrade note: after deploying 0.11.0, add standards = "~/.claude/skills/larv-review/standards" to each config under ~/.claude-projects/, as the changelog says.

Release review of 0.11.0: the architecture and principles passes over the whole project, and the fixes for the drift they found. **What changed** - The plan's packages import one way: a pure `bin/plan/` core, `bin/atlassian/` for the Jira and Confluence client, and `plan_store` split into `plans`, `todo_md`, `scoring` and `stores`. A repo check fails on any cycle among `bin/`'s packages. - One parse and one print of an item's cells, `item_fields` and `row_cells`, for every plan backend. - A pure `room_of` decides which chunks start; the settings are a table, `SCALARS`; `Standing` weighs the plan against the cut once; the next chunk's pick is `release_plan/next_chunk.py`'s `NextChunk`; loop-claude reads the plan once per cycle start. - chunk-pr loads the config in-process, and the title rule has one home, `templates.misfits`. - Record stores keep no hidden version: `held()` and `write(text, over)`. lilleman chose this on 2026-11-16, and the decision is reworded to match. - rule-hook's `run-heavy` rule also catches runs through `uv run`, `poetry run`, `sudo` and `doas`. lilleman chose this on 2026-11-16. - The notifier backends live in `notifier/`, and one definition covers the changelog's unreleased heading. - lilleman's L 3 / L 4 routing ruling is filed as item 393 for 0.12.0, by their choice. The regrouping of `bin/` is filed as item 392 for 0.14.0. **Reviews** - Architecture: 3 rounds. BLOCK in rounds 1 to 3; lilleman confirmed the restructure, and each round's findings are fixed. - Principles: 2 rounds. Its findings are fixed, filed or declined below. - Security: `/security-review` found nothing. The full `claude-security` scan could not run in a subagent. - Overrides: 2 rounds, approved. - Stability: 1 round, 14 🟡, no 🔴; all are fixed, three of them as places, below. - Product-owner: 1 round, 2 🟡, fixed. - Prose: 1 round, 9 🟡, fixed. - No auto-reviewer is configured. **Comprehension check**: paired ruling **better** at 31da587 against 64454f8. Mid A voted Navigation, Locality and Shape better, Self-sufficiency same, Overall better. The inherited architect voted better on all four and overall. No scoring run was due. **Overrides**: `todo-rank-check` on the commits that changed `todo.md`. Reason: the branch's own `rank_todo.py` needs 0.11.0's required `standards` setting, which this machine's config can't hold until 0.11.0 is deployed, since v0.10.1's tools refuse it. Both the released and the branch's checker passed. Ruling: approved; the overrides on empty commits are moot. **Places left as they were, and why**: `loop_claude/cycle/run.py` `rule`, `loop_claude/cycle/judge.py` `INTERRUPT_RULES` and `larv_review/plan.py` `scoring_run_due` each need a restructure outside this release's themes. Floor item 300 holds them. **🟢 declined** - OTLP traces: loop-claude's archive is this CLI setup's trace of each cycle. Recording it as a decision would need an item that ends the departure, which is lilleman's call. - The two `Http` aliases are each used only inside their own module. - `Scalar.read` is typed `Any`, since a mixed table can't tie each reader's type to its field. - deploy-release and adf-codec print progress from their imperative shells. - chunk-pr's size is item 287's to cut. - `adf_to_markdown` is the codec's one-document API. - Item 300's quoted places are left for its next refresh. **Upgrade note**: after deploying 0.11.0, add `standards = "~/.claude/skills/larv-review/standards"` to each config under `~/.claude-projects/`, as the changelog says.
lilleman added 1 commit 2026-11-16 13:56:12 +01:00
Release review of 0.11.0
CI / tests (pull_request) Successful in 2m47s
3f7d85c48e
lilleman added 5 commits 2026-11-16 16:12:47 +01:00
lilleman added 4 commits 2026-11-16 16:24:10 +01:00
lilleman added 3 commits 2026-11-16 16:53:41 +01:00
lilleman marked the pull request as ready for review 2026-11-16 16:53:48 +01:00
lilleman added 2 commits 2026-11-16 16:57:48 +01:00
lilleman merged commit 8d24f73971 into main 2026-11-16 17:01:08 +01:00
Sign in to join this conversation.
No Reviewers
No Label
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: lilleman/claude-config#562