v4.1.0: per-task reviewer + per-role model assignments - #60
Merged
Merged
Conversation
Add Goal and Verification lines to the /code:fix Task Brief template, and a Rules entry making them mandatory. Closes the "did we achieve it" loop for the bug lane symmetrically with the feature lane (where /code:plan + /code:ship already enforce expected_outcome + verification end-to-end). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add a per-task review pass between implementer-returns and merge: capture diff, dispatch a Sonnet 4.6 reviewer fork that calls Skill("code-review") (with a 5-area inline fallback mirroring /code:review), and on CRITICAL/HIGH findings dispatch one review fix-pass before merging into the parent feature branch. Same one-cycle retry budget as the post-merge audit.
Set explicit model per role across the swarm: Opus 4.7 for orchestrator + audit fix-pass (judgment), Sonnet 4.6 for implementer + reviewer + review fix-pass (routine coding), Haiku 4.5 for parallel Explore in /code:plan (breadth).
Document the convention in code-et-implementer/CLAUDE.md so it's discoverable in any installed project. Subsumes the v4.0.1 fix.md Goal+Verification work that was committed earlier on this branch.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
/code:ship— capture diff after each implementer subagent commits, dispatch a Sonnet 4.6 reviewer fork that callsSkill("code-review")(with a 5-area inline fallback mirroring/code:review), and on CRITICAL/HIGH findings dispatch one review fix-pass before merging into the parent feature branch. One-cycle retry budget, same as the post-merge audit./code:plan(breadth). Documented incode-et-implementer/CLAUDE.md./code:fixTask Brief now states Goal + Verification (subsumed from interim 4.0.1 commit) — closes the "did we achieve it" loop for the bug lane symmetrically with the feature lane, where/code:planalready encodesexpected_outcome+verificationin TaskCreate metadata.Test plan
/code:fix "<sample bug>"in a scratch Rust repo — confirm the produced Task Brief contains the new**Goal:**and**Verification:**lines./code:plan "<sample feature>"followed by/code:shipin a scratch repo with the engineering plugin installed — confirm each task triggers a per-task reviewer fork (Sonnet 4.6), and that a planted CRITICAL finding (e.g., layer violation) triggers exactly one review fix-pass before merge.Agentmodel overrides resolve correctly: implementer + reviewer subagents run on Sonnet, audit fix-pass runs on Opus, parallel Explore in/code:planruns on Haiku./code:reviewstill runs cleanly post-merge (per-task review is shift-left, doesn't replace the pre-PR gate).🤖 Generated with Claude Code