← ClaudeAtlas

vibe.verifylisted

승인된 SPEC에 대한 구현 완료를 판정하거나 REQ→SPEC→Feature→Code 추적표가 필요할 때 검증 또는 trace mode로 실행한다.
su-record/vibe · ★ 0 · AI & Automation · score 75
Install: claude install-skill su-record/vibe
# /vibe.verify ## 완료 기준 - [ ] 승인된 SPEC의 모든 REQ가 pass 또는 fail로 판정되어 있다. - [ ] 각 판정에 test output, 파일·행 또는 산출물 경로 근거가 있다. - [ ] verify mode 성공 시 run ledger의 `verifyPassed`가 true다. - [ ] trace mode에서는 RTM 파일이 존재하고 누락 REQ가 0개다. SPEC-driven verification. Check the implementation against the SPEC's **Done criteria** and Feature scenarios, record the result to the **run ledger** (`recordVerify`), and auto-register regressions on failure. The JUDGE gate is code-enforced by the ledger — a verification never counts as passed by self-report. ## Usage ``` /vibe.verify "feature-name" # SPEC-based verification /vibe.verify --e2e "feature-name" # + E2E browser verification (agents/e2e-tester.md) /vibe.verify trace "feature-name" # Requirements Traceability Matrix /vibe.verify trace "feature-name" --html | --save | --json ``` ## Trace mode When the first argument is `trace`, read `references/trace-mode.md` and execute that RTM workflow instead of the verification process below. Preserve the feature name and `--html`, `--save`, and `--json` flags. An empty RTM is a gate failure. Trace output alone does not set `verifyPassed`; run normal verify mode to bind command evidence to the run ledger. ## Scope - **Changed files only** — never full-project scan. - Verification target = the SPEC's Done criteria + Given/When/Then scenarios. Read whatever is needed to judge each criterion accurately; choose your own reading strategy, but never mark a criterion passed without concr