此技能识别 GitHub Agentic Workflows 相关请求的类型,并引导代理读取对应提示或技能文件,涵盖设计、创建、配置、更新、调试和升级。
- 设计或创建新工作流
- 更新现有工作流
- 调查工作流错误
使用 `gh skill` 搜索和预览 Agent Skills,指定代理与保存范围进行安装,更新已安装的技能,并检查仓库布局和 `SKILL.md` 元数据,准备在 GitHub 上发布可搜索的技能来源。
Manage agent skills with gh skill. Use this skill to discover, preview, install, update, and publish Agent Skills so an agent can self-manage the skills available in its environment.
实测评分 未实测
查看安装方法使用 gh skill 搜索 GitHub 上的 Agent Skills,查看技能内容,再进行安装和更新。安装时可以指定目标代理,以及将技能保存到当前项目还是用户范围。也可以将安装版本固定到标签、分支或提交,或从本地仓库安装。准备发布时,此技能会检查仓库中的技能目录结构和 SKILL.md 元数据,并介绍如何将仓库作为可搜索的技能来源发布到 GitHub。
想从 GitHub 寻找技能,并在安装前检查内容时,可以使用它。需要在多个代理环境中使用不同技能、集中更新已安装的技能,或整理自己的技能仓库并准备发布时,也适用。
安装时需要指定目标代理。发布操作会涉及 GitHub 仓库,并且发布流程需要指定标签。
本介绍由 AI(OpenAI Codex)根据公开的 SKILL.md 撰写。准确的规格请查看下方原文。
AI 根据 SKILL.md 的内容检查了需要注意的操作和使用前提。评分只依据实际运行的结果(实测对比)。
不计入评分。 由 AI(OpenAI Codex)于 2026年10月26日 评估。仅供参考,安装前请同时查看原文。
安装前可以查看交给 AI 代理的指令内容。
gh skillgh skill installs, previews, searches, updates, and publishes
Agent Skills. An agent can use it to keep its
own skill set in sync with one or more GitHub repositories.
The command is also aliased as gh skills. Prefer the canonical singular
gh skill in scripts and docs.
gh skill search <query> # free-text search
gh skill search <query> --owner <org> # restrict to one owner
gh skill search <query> --limit 20 --page 2
gh skill search <query> --json skillName,repo,description
gh skill preview <owner>/<repo> <skill-name>
gh skill preview <owner>/<repo> <skill-name>@v1.2.0 # pin a version
gh skill install <owner>/<repo> <skill-name>
gh skill install <owner>/<repo> <skill-name>@v1.2.0
gh skill install <owner>/<repo> skills/<scope>/<skill-name> # exact path, fastest
gh skill install ./local-skills-repo --from-local
<owner>/<repo> and <skill-name> are both required.
Useful flags:
--agent <id> - target host (e.g. github-copilot, claude-code,
cursor, codex, gemini-cli). Repeat for multiple. Default is
github-copilot when non-interactive. You should know what agent you are,
so set this appropriately to install for yourself.--scope project|user - project (default) writes inside the current
git repo; user writes to the home directory and applies everywhere.--pin <ref> - pin to a tag, branch, or commit SHA. Mutually exclusive
with --from-local and with inline @version syntax.--allow-hidden-dirs - also discover skills under dot-directories such
as .claude/skills/. Don't use this unless you need to, it comes with risks.--force - overwrite an existing install.gh skill update --all # update every installed skill
gh skill update <skill> # update one
gh skill update <skill> --force
gh skill update --unpin # drop the pin and move to latest
Publishing turns a repo into a discoverable skill source. Skills are discovered with these conventions:
skills/<name>/SKILL.mdskills/<scope>/<name>/SKILL.md<name>/SKILL.md (root-level)plugins/<scope>/skills/<name>/SKILL.mdEach SKILL.md needs YAML frontmatter:
---
name: my-skill # must equal the directory name
description: One sentence... # required, recommended <= 1024 chars
license: MIT # optional but recommended
---
gh skill publish --dry-run # validate only, no release
gh skill publish --dry-run ./path/to/repo # validate a specific dir
gh skill publish --fix # auto-strip install metadata
gh skill publish --tag v1.0.0 # non-interactive publish
gh skill publish # interactive publish flow
--fix and --dry-run are mutually exclusive. --fix only rewrites
install-injected metadata.github-* keys and does not publish; commit
the result and re-run publish.
The publish flow will:
agent-skills topic to the repo (so search can find it).--tag (or prompt for one in a TTY).Always pass --tag so it doesn't fall through to the interactive flow.
A reasonable loop:
gh skill search <topic> --json skillName,repo,namespacegh skill preview <repo> <skill> to inspect the SKILL.md.gh skill install <repo> <skill> --agent <host> --pin <ref> for a
reproducible install.gh skill update --all to refresh.还没有评价。欢迎分享你的使用感受。
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。
此技能识别 GitHub Agentic Workflows 相关请求的类型,并引导代理读取对应提示或技能文件,涵盖设计、创建、配置、更新、调试和升级。
Exercise a real command-line program with specified cases or a stated goal, then capture and verify the observed behavior. It produces local media evidence such as a session video, helping reviewers see what the command did without treating the recording as proof of general correctness.
中文简介正在准备中。当前显示英语简介。
将命令行练习的预期与记录证据逐项核对,按需渲染并检查媒体,返回各案例结果和本地成果文件路径;默认不发布证据。
此技能按原样执行指定的命令行测试,保留命令、输入、顺序和预期条件;逐项准备、操作并验证,再依据实际证据标记通过、失败、受阻或仅作观察。
此技能根据目标和已批准的范围选择可观察的命令行检查,逐项完成准备、操作和即时验证,并整理证据、尚未达到的行为及检查结果。
这是供 `cli-exercise` 内部使用的准备检查模块,用于核验构建、终端和媒体工具是否可用;缺少依赖时先征得安装同意,安装后重新检查。