review-workflow

name	review-workflow
description	The implement→review flow under the ≤3-reviewer budget with the conditional test stage. Use for /implement-feature, /fix-bug, /refactor-module and any change that needs structured review.

Review workflow

1. Analyze the request; identify affected files (graphify/codegraph first, not repo scans). Pre-harvest the paths + signatures into a handoff file .claude/scratch/<task>.md (copy HANDOFF_TEMPLATE.md) so no stage cold-re-explores.
2. Select the pair by Ownership precedence: the area's author (implements) + its independent critic, plus a Secondary critic only if the change is risk-bearing in that column (tauri-security-reviewer, performance-profiler, ui-ux-expert). ≤3 critics total.
3. Plan the minimal change (in the handoff).
4. Implement — the domain author makes minimal changes (reads the handoff first; loads author-contract). Rust-first for business logic / pipelines / ATS / document generation; the renderer stays presentation-focused. The author appends what changed to the handoff.
5. Test stage (conditional) — if touchesTestableLogic(diff) (Part D predicate): test-author writes/updates tests → testing-reviewer audits coverage of the changed code. This stage is separate from the ≤3-critic cap. STRICT: changed non-trivial logic without a test — or a weak/tautological/mock-asserting test that doesn't exercise the change — is HIGH (blocking), not advisory; the error/edge path must be covered, not just the happy path.
6. Review — the independent critic (never the author) audits the diff against its checklist and appends severity-tagged findings to the handoff; HIGH/CRITICAL block → author resolves → re-audit; LOW/MEDIUM advisory. For genuinely parallel, file-disjoint, multi-domain work, run this as native Agent Teams (the lead spawns authors as teammates owning disjoint files; critics challenge via the mailbox); otherwise sequential subagents (cheaper).
7. Verify correctness — run the relevant tests/build (rtk pnpm test, cargo test).
8. Verify performance — if a hot path was touched → performance-profiler.
9. Verify security — if risk-bearing → tauri-security-reviewer (HIGH/CRITICAL blocks).
10. Docs + lessons — project-steward syncs affected docs/knowledge, runs graphify update ., and persists any durable lesson.

No feature is "done" without tests (when the predicate is positive) or without docs sync.