跳到正文
cccskills
Codex 官方免费在 GitHub 公开

code-change-verification

Select and run Guardrails verification for code, dependency, test, tooling, documentation, or repository-workflow changes.

中文简介正在准备中。当前显示原文说明。

查看安装方法

包含的文件(2)

  • SKILL.md2.4 KB
  • agents/openai.yaml123 B

SKILL.md(原文)

安装前可以查看交给 AI 代理的指令内容。

Code Change Verification

Follow repository verification and inspect the current package.json scripts before selecting checks. Run from the selected linked worktree's root; preserve its branch and user-owned changes.

Select the evidence

Changed areaVerification
Runtime, public types, dependencies, build/test configuration, packagingFull sequence below; focused tests during implementation
SDK integration or dependency compatibilityBuild first, then the relevant src/__tests__/integration/sdk-*.test.ts tests; full sequence before handoff
Site content, links, navigation, or VitePress toolingnpm ci, then npm run docs:check
Contributor instructions or skill metadata onlyValidate commands, relative links, frontmatter/UI metadata, and git diff --check

When areas overlap, combine their required checks. A test described as "integration" is not necessarily live: inspect its fixtures and calls before running it. The normal suite uses local fixtures and mocks; do not turn ordinary verification into a credentialed eval or an external API experiment.

Full sequence

Run each command separately and inspect its exit status before proceeding:

npm ci
npm run build
npm run test:run
npm run lint
npm run docs:check

The build emits CommonJS and declarations consumed by compatibility tests. test:run avoids watch mode. Biome lint includes formatting and import checks; VitePress and the Node documentation tests run through docs:check. Invoke npm directly in the host shell, including PowerShell; no platform-specific wrapper or additional formatter is needed.

Keep verification within the available sandbox. Diagnose failures and fix only in-scope defects; report unrelated failures and missing coverage. Retry a transient failure when evidence supports that diagnosis. After a fix, repeat checks whose evidence it invalidated. Never report a skipped or failed gate as passing, and do not infer Windows coverage from Linux or macOS results.

Report the command, result, tested Node version, and any coverage limits. For remote checks, associate the result with the current PR head. The review gate in AGENTS.md also applies before pushing; passing tests do not replace it.

评价

还没有评价。欢迎分享你的使用感受。

评论

无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。

还没有评论。来写第一条吧。

提交后会立即公开。请勿填写邮箱、电话等个人信息。URL 最多 2 个。

0 / 1000 字

此表单受 reCAPTCHA 保护,适用 Google 的隐私权政策和服务条款。

同一仓库的技能

implementation-final-review

Codex 官方免费

Prepare independent review of a complete Guardrails change before pushing or updating a PR, using the repository adversarial-review procedure.

中文简介正在准备中。当前显示原文说明。

openai/openai-guardrails-js1102026年10月21日 更新

implementation-kickoff

Codex 官方免费

Start or resume a requested Guardrails implementation or PR takeover in the selected linked worktree with bounded scope and verification.

中文简介正在准备中。当前显示原文说明。

openai/openai-guardrails-js1102026年10月21日 更新

implementation-strategy

Codex 官方免费

Choose bounded implementation scope and existing owning modules before Guardrails runtime, configuration, client, resource, streaming, Agents, or SDK compatibility changes and feedback fixes.

中文简介正在准备中。当前显示原文说明。

openai/openai-guardrails-js1102026年10月21日 更新

pr-draft-summary

Codex 官方免费

Draft a Guardrails PR from its complete final diff and carry out authorized CI and review follow-up, including stacked PRs and takeovers.

中文简介正在准备中。当前显示原文说明。

openai/openai-guardrails-js1102026年10月21日 更新

查看 openai 的全部技能

报告此技能的问题