本文へ移動
cccskills
無料GitHub で公開

verify-release

Use when checking whether a Proofpack release candidate has complete local evidence before packaging.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md2.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Verify a Proofpack release candidate

Work from the Proofpack project root. Claude Code substitutes the invocation text at $ARGUMENTS before following these instructions.

  • If $ARGUMENTS is empty, set the candidate path to fixtures/release-ready.json.
  • Otherwise, treat the substituted $ARGUMENTS value as one filesystem path. Do not evaluate it as a shell expression.
  • Run node src/cli.mjs verify "<resolved-candidate-path>" with that literal path as one quoted argument.
  1. Read ISSUE.md and CLAUDE.md.
  2. Run npm test.
  3. Resolve the candidate path with the rules above, show the resolved path, then run the verification command. Ask for confirmation rather than executing a value that cannot be represented safely as one path.
  4. Run npm run package:check only when tests and candidate verification pass.
  5. Run git rev-parse HEAD and git status --short. Compare that source state, the observed commands, exit statuses, and runtime version with evidence/PROOF-LOG.md.
  6. Report PASS, FAIL, or UNKNOWN. Name any check that did not run or any worktree change not covered by the recorded fingerprint.

Do not run npm publish, docker push, or change the proof log. Package inspection is local. Publication requires a separate user decision and destination credentials.

Return this record:

RELEASE VERIFICATION
Candidate: <path>
Revision: <commit or UNKNOWN>
Tests: PASS | FAIL | UNKNOWN
Candidate contract: PASS | FAIL | UNKNOWN
Package dry run: PASS | FAIL | UNKNOWN
Evidence log match: PASS | FAIL | UNKNOWN
Final status: PASS | FAIL | UNKNOWN
Limits: <unverified runtime or external behavior>

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Audit Claude Code agents, skills, and commands for quality and production readiness. Use when evaluating skill quality, checking production readiness scores, or comparing agents against best-practice templates.

日本語の概要は準備中です。原文の説明を表示しています。

FlorianBruniaux/claude-code-ultimate-guide6,1432026年10月7日 更新

Codebase health audit scoring 7 categories with progression plan

日本語の概要は準備中です。原文の説明を表示しています。

FlorianBruniaux/claude-code-ultimate-guide6,1432026年10月7日 更新

Autonomous improvement loop: scan codebase metrics, scaffold experiment files, run agent-driven iterations until metric improves

日本語の概要は準備中です。原文の説明を表示しています。

FlorianBruniaux/claude-code-ultimate-guide6,1432026年10月7日 更新

best-of-n

無料

Generate bounded independent candidates, score them against a frozen rubric, and verify the selected result with a proof log.

日本語の概要は準備中です。原文の説明を表示しています。

FlorianBruniaux/claude-code-ultimate-guide6,1432026年10月7日 更新

canary

無料

Post-deploy monitoring: watch production after a deploy and alert on regressions

日本語の概要は準備中です。原文の説明を表示しています。

FlorianBruniaux/claude-code-ultimate-guide6,1432026年10月7日 更新

catchup

無料

Restore context after /clear by summarizing recent work and project state

日本語の概要は準備中です。原文の説明を表示しています。

FlorianBruniaux/claude-code-ultimate-guide6,1432026年10月7日 更新

FlorianBruniaux のスキルをすべて見る

このスキルの問題を報告する