メニュー
Pre-commit verification. Use when about to commit, asked to "verify", "검증", "커밋 전 확인", or before any git commit.
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
Deploy to Mac Mini production server. Use when asked to "deploy", "배포", "push to production", or "서버 반영".
STRATEGY §2.7 7-phase Flow (Think→Plan→Build→Review→Test→Ship→Reflect) orchestrator. **Recommend-only** — 자동 chain 없이 사용자 현재 phase 추정 후 다음 액션/스킬 권고. 사용자가 "다음 단계 뭐?", "지금 phase 가 뭐야?", "Reflect 해야 하나?", "Build 끝났는데 뭘 해야 해?" 같이 진행 단계 질문 시 발화. nuri-quant 프로젝트의 1 issue = 1 PR ≤ 3 commits 룰 준수 보조.
Code review checklist. Use when asked to "review", "리뷰", "코드 검토", or when reviewing a PR diff.
SIEGE predictivity measurement (E4-0b v2) audit 재실행 + 해석 절차. Use when re-running 60-month predictivity audit, interpreting gate-level Δ results, or adding new variant templates. audit route 는 2026-04-22 currently closed state — 재활성화 조건 확인용.
LLM 에이전트 실패 패턴 디버깅 — 코드/툴 레벨 (§5.1-5.6: hallucination, phantom fix, test illusion, scope creep, context bias, stale number drift) + 대화/추론 레벨 (§5.14.1-2: data→recommendation slide, cross-context inconsistency). Use when user reports "test passes but doesn't cover the code", "fix applied but bug persists", "why does this keep failing the same way", "Claude is hallucinating API", "갑자기 추천한 이유가 뭔가요", "왜 KR/US 다르게 적용했나요", or when a regression test needs a Gotcha-Test Pair citation. Case study narrative: git log of PR #272, #300-#307, session 2026-05-27.
| name | nuri-verify |
| description | Pre-commit verification. Use when about to commit, asked to "verify", "검증", "커밋 전 확인", or before any git commit. |
make verify-quick
make verify-all
make lint passes (ruff check)make test passes (6,210 tests, xdist parallel)grep -ri "old_value" across CLAUDE.md, README.md, STRATEGY.mdconfig/rules.yaml or config/agents.yaml, not hardcodeddatetime.now() 사용? → kst_now() 또는 today_kst()로 교체nuri/core/db/ 함수로 교체_MIGRATIONS 리스트에 추가(feat|fix|docs|refactor|test|chore)(...): ...