メニュー
Use when turning the github-pr-review-comments JSONL index, including CodeRabbit PR review comments captured there, into an action queue with must-fix, verify-first, duplicate, informational, and human-decision buckets.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Use when managing per-project backlog folders, feature queues, async agent work items, lifecycle status updates, PR/done/blocked transitions, or status reviews inside an Obsidian vault.
Use when prototyping, authoring agent skills, or asking the agent to zoom out and map unfamiliar code before acting.
Create new agent skills with proper structure, progressive disclosure, and bundled resources. Use when user wants to create, write, or build a new skill.
Use when reviewing architecture boundaries, hexagonal architecture, clean architecture layers, dependency direction, trust boundaries, uploads, parsers, redirects, or web security risks.
Use when configuring or running hexcheck in a Go repo for hexagonal architecture boundaries, adapter business-logic warnings, and mock discipline.
Find deepening opportunities in a codebase, informed by the domain language in CONTEXT.md and the decisions in docs/adr/. Use when the user wants to improve architecture, find refactoring opportunities, consolidate tightly-coupled modules, or make a codebase more testable and AI-navigable.
| name | github-pr-finding-triage |
| description | Use when turning the github-pr-review-comments JSONL index, including CodeRabbit PR review comments captured there, into an action queue with must-fix, verify-first, duplicate, informational, and human-decision buckets. |
Use after github-pr-review-comments has produced a commentIndexJsonl file.
This is classification of reviewer claims, not proof that the claims are valid. Verify code before changing behavior.
Do not downgrade refactoring, cleanup, or maintainability findings merely because they are inconvenient or behavior still works; apply code-review-release/subskills/thermo-nuclear-code-quality-review/SKILL.md before dismissing them.
# Build step (one-time or when tools change):
tsc --project $HOME/.agents/skills/github/tools/tsconfig.json
# Then run:
node $HOME/.agents/skills/github/tools/dist/pr-finding-triage.js --index /tmp/...comments.jsonl --repo OWNER/REPO --pr PR_NUMBER
Outputs:
/tmp/tmpmust_fix: reviewer text claims critical/major/security/failing/broken behavior.verify_first: inline or uncertain findings that need local code verification before action.probably_duplicate: same normalized path/line/severity/body preview as an earlier finding.informational: nitpick/trivial/info/style-only comments. Not for structural maintainability findings unless verified against the thermo-nuclear quality bar.needs_human_decision: API/product/UX/tradeoff or low-confidence comments.Each finding includes:
derivedSeverityconfidenceclassificationBasisrationalepr-review-comments and confirm it is complete (ok: true, no pagination gaps).must_fix, then verify_first, then needs_human_decision.probably_duplicate without checking duplicateOfId and context.pr-review-verification before implementing fixes.The classifier is heuristic. It prioritizes actionability and safety over cleverness. A must_fix means “reviewer appears to claim this is important,” not “the code is definitely wrong.”