菜单
Scan .agents/skills/ and generate matching command wrappers in .agents/commands/. Use when adding a new skill, renaming a skill, or auditing that every skill has a corresponding command. Also use when a skill directory structure changes (e.g. flat file to directory with SKILL.md).
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Run pnpm commands for this monorepo. Use this skill whenever the user asks to install dependencies, add/remove/update packages, run dev/build/test scripts, or manage the pnpm workspace — whether they say "pnpm", "装包", "加依赖", "更新依赖", "add dependency", "install package", "run script", "dev", "build", "clean", or any package-management operation. Also use it when the user is unsure which command to use or what a root script does.
Run the right local and CI-aligned validation checks for this pnpm monorepo. Use this skill whenever the user asks to test changes, run unit tests, run related e2e tests, verify a branch locally, prepare for PR/CI, or says phrases like "跑测试", "本地验证", "unit/e2e", "相关 e2e", "test check", "CI 检查", "改完跑一下测试", "不要全量 e2e", or asks which checks should be run. For quick local validation, run root unit tests and only related Playwright e2e specs; for PR/CI confidence, align with .github/workflows/ci.yml: lint, E2E coverage, typecheck, build, package artifact validation, unit tests, and related E2E specs unless full E2E is explicitly requested or clearly necessary.
Full git workflow from unstaged changes to a merged PR. Use this skill whenever the user wants to turn current changes into a GitHub issue, proper branch, commit, quality gate, and pull request — especially when they say things like "提交并提 PR", "create an issue and PR for these changes", "按照改动建 issue 提 pr", "push these changes with a proper issue", or any request that involves creating an issue and PR from working-tree changes. This skill orchestrates the tool-git-issues, tool-git-commit, tool-test-check, and CI skills into a single workflow, and it must bring in std-antfu-vue when PR changes touch Vue SFCs, theme Vue components, NewLayout.vue, Composition API, props/emits, composables, or Vue UI refactors.
Use this coding-standard skill for Vue work in this VitePress Theme Link repo, especially when editing .vue SFCs, NewLayout.vue, theme components, Composition API logic, props/emits contracts, composables, UI behavior, or Vue refactors. It adapts antfu/skills@vue-best-practices: keep the full reference-driven workflow, but preserve this repo's existing Prettier/ESLint setup, avoid unnecessary abstraction, and require related/full E2E validation for UI changes.
Control a headless Playwright Chromium browser to inspect live DOM, CSS computed styles, take screenshots, and evaluate JavaScript on any URL. Use this skill whenever the agent needs to verify VitePress SSR/hydration, inspect UnoCSS generated classes, debug plugin client-side behavior, check responsive layout or dark mode, extract DOM/CSS design tokens from external websites (GitHub, blogs, etc.), or answer questions about how a page actually renders. Triggers on: "inspect the DOM", "check the rendered HTML", "take a screenshot", "get computed styles", "what CSS does this use", "how does this page render", "browser inspect", "playwright explore", "evaluate in browser", "open in browser", "看看这个页面的样式", "截图", "检查 DOM", "浏览器看一下", and any request to extract live page information that static file reading cannot provide.
Script-driven synchronization between local Markdown issues under .ai_dev/issues and a GitHub Projects Kanban board using gh project / GraphQL. Use this skill whenever the user asks to sync kanban, GitHub Projects, project board columns, .ai_dev/issues, local issue files, pull remote issues into local markdown, push local issues to the project board, or map local issue status to remote Kanban status. Prefer the bundled scripts over ad-hoc shell commands: pull means remote overwrites local after human confirmation, push means local overwrites remote after reviewing the generated plan.
基于 SOC 职业分类
| name | self-sync-skills-command |
| description | Scan .agents/skills/ and generate matching command wrappers in .agents/commands/. Use when adding a new skill, renaming a skill, or auditing that every skill has a corresponding command. Also use when a skill directory structure changes (e.g. flat file to directory with SKILL.md). |
Ensure every skill under .agents/skills/ has a matching command entry in .agents/commands/.
.agents/skills/ for directories that contain a SKILL.md.<name>, check whether .agents/commands/<name>.md exists.../skills/<name>/SKILL.md reference path is still correct (update if stale)../scripts/link_agents.sh to propagate symlinks to IDE directories.Use this exact template for every generated command:
---
name: <skill-name>
description: '<one-line description>'
---
Invoke the skill at `../skills/<skill-name>/SKILL.md` and follow its instructions exactly.
<skill-name> and <one-line description> from the corresponding SKILL.md frontmatter (name and description fields).../skills/<name>/SKILL.md works uniformly from every IDE commands/ directory after link_agents.sh creates the per-item symlinks..agents/skills/ that does not contain a SKILL.md file (it is not a skill).self-skill-creator is itself a skill and should have a command just like any other.self-sync-skills-command itself also needs a command — treat it like any other skill../scripts/link_agents.sh so IDE directories pick up the changes.