ワンクリックで
verify-slice
Run the active Project's build and typecheck heuristic and return a structured verification result.
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。
メニュー
Run the active Project's build and typecheck heuristic and return a structured verification result.
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
Implement one named Project issue with red-green TDD and return a structured task outcome.
Use when writing or reviewing TypeScript/full-stack code. Encodes principles for type safety (branded types, discriminated unions, end-to-end types), real tests over mocks, OpenTelemetry observability, and picking the right abstractions instead of premature ones.
Use when designing, writing, reviewing, or refactoring React components and features. Covers component boundaries, state ownership, composition, prop design, custom hooks, useEffect avoidance, memoization, Zustand, TanStack Query, loading/error/empty states, async event handling, TypeScript state modeling, and file organization. Trigger for React component design questions, component-tree restructuring, state management choices, data fetching, effect cleanup, reusable UI composition, prop drilling, render performance, error boundaries, optimistic updates, double-submit prevention, race conditions, or organizing a React codebase.
Generate Brief artifact content and persist it through the Brief and Project context tools.
Use only after a fresh slice PRD exists under the Project's .cairn/prds/ folder and the user-approved slice needs to be broken into local implementation issue files.
Use only after the project brief is concrete, the Cairn has proposed a smallest meaningful first slice, and the user has agreed to that slice; silently draft the slice PRD.
| name | verify-slice |
| description | Run the active Project's build and typecheck heuristic and return a structured verification result. |
| disable-model-invocation | false |
| response_schema | verify_result |
| args | {"type":"object","required":["project_root"],"properties":{"project_root":{"type":"string","description":"Absolute path to the active Project root."}}} |
Run the active Project's verification checks silently and return a structured result.
Inspect <project>/package.json.
If there is no package.json, or if it has no usable scripts, verification succeeds:
{ "ok": true, "message": "No build or typecheck script configured." }
If scripts.build is a string, run:
bun run build
If scripts.typecheck is a string, run:
bun run typecheck
Run build first, then typecheck. Stop at the first failing check.
Return only one JSON object with this exact shape:
{ "ok": true, "message": "Verification passed." }
Use { "ok": true, "message": "Verification passed." } only when every configured check exits with code 0.
Use { "ok": false, "message": "..." } when a check cannot run or exits non-zero. Keep the message short and factual. Include the script name and the most useful stderr or stdout excerpt, trimmed to the essential failure.
Do not modify Project files. Do not add tests. Do not call a real LLM.