@@ -102,122 +102,14 @@ pre-agent-steps:
102102 skill_name="${skill_part%%@*}"
103103 pin_ref="${skill_part#*@}"
104104 echo "Installing skill: $repo $skill_name (pinned to $pin_ref)"
105- gh skill install "$repo" "$skill_name" --agent "$agent" --pin "$pin_ref" $upstream_flag
105+ gh skill install "$repo" "$skill_name" --agent "$agent" --pin "$pin_ref" --allow-hidden-dirs --force $upstream_flag
106106 else
107107 echo "Installing skill: $repo $skill_part"
108- gh skill install "$repo" "$skill_part" --agent "$agent" $upstream_flag
108+ gh skill install "$repo" "$skill_part" --agent "$agent" --allow-hidden-dirs --force $upstream_flag
109109 fi
110110 else
111111 echo "Installing all skills from: $repo"
112- gh skill install "$repo" --agent "$agent" $upstream_flag
112+ gh skill install "$repo" --agent "$agent" --allow-hidden-dirs --force $upstream_flag
113113 fi
114114 done < <(echo "$skills_json" | jq -r '.[]')
115115---
116-
117- <!--
118- ## Agent Skills
119-
120- This shared workflow installs agent skills before the AI agent runs, using the
121- [`gh skill install`](https://cli.github.com/manual/gh_skill_install) command
122- (available in GitHub CLI v2.90.0+).
123-
124- ### How it works
125-
126- Each skill in the `skills:` list is installed into the repository's
127- `.github/skills/` directory via `gh skill install`, making the skills available
128- to the AI agent during its session.
129-
130- ### Usage
131-
132- ```yaml
133- engine: copilot
134- imports:
135- - uses: shared/gh-skill.md
136- with:
137- skills:
138- - github/awesome-copilot/documentation-writer
139- - github/awesome-copilot/code-review@v1.2.0
140- ```
141-
142- For other engines, set `engine:` in the `with:` block to match your workflow
143- engine so skills are installed for the correct agent host:
144-
145- ```yaml
146- engine: claude
147- imports:
148- - uses: shared/gh-skill.md
149- with:
150- engine: claude
151- skills:
152- - github/awesome-copilot/documentation-writer
153- ```
154-
155- To opt out of the upstream update:
156-
157- ```yaml
158- imports:
159- - uses: shared/gh-skill.md
160- with:
161- upstream: false
162- skills:
163- - github/awesome-copilot/documentation-writer
164- ```
165-
166- ### Inputs
167-
168- | Input | Required | Description |
169- |-------|----------|-------------|
170- | `skills` | ✅ | List of skills to install (see formats below) |
171- | `engine` | No | gh-aw engine name — determines the `--agent` target (default: `copilot`) |
172- | `github-token` | No | GitHub token for downloading skills (default: built-in `GITHUB_TOKEN`) |
173- | `upstream` | No | Pass `false` to skip `--upstream` and use the cached version (default: `true`) |
174-
175- ### Skill format
176-
177- Each entry in `skills:` is one of:
178-
179- | Format | Example | Effect |
180- |--------|---------|--------|
181- | `owner/repo` | `github/awesome-copilot` | Installs all skills from the repo |
182- | `owner/repo/skill-name` | `github/awesome-copilot/documentation-writer` | Installs a specific skill (latest, with `--upstream`) |
183- | `owner/repo/skill-name@ref` | `github/awesome-copilot/code-review@v1.2.0` | Installs with `--pin ref` |
184-
185- The `@ref` part is extracted from the skill path and passed as `--pin ref` to
186- `gh skill install`. Any git ref is accepted (tag, branch, SHA).
187-
188- ### Engine → agent mapping
189-
190- | `engine` input | `gh skill --agent` value |
191- |---------------|--------------------------|
192- | `copilot` (default) | `github-copilot` |
193- | `claude` | `claude-code` |
194- | `codex` | `codex` |
195- | `gemini` | `gemini-cli` |
196- | `opencode` | `opencode` |
197-
198- ### Authentication
199-
200- Uses the token provided via the `token` input, falling back to the built-in
201- `GITHUB_TOKEN`. For private skill repositories, pass a token with read access:
202-
203- ```yaml
204- imports:
205- - uses: shared/gh-skill.md
206- with:
207- github-token: ${{ secrets.MY_SKILLS_TOKEN }}
208- skills:
209- - my-org/private-skills/my-skill
210- ```
211-
212- ### Network access
213-
214- `gh skill install` downloads skill files from GitHub. No additional network
215- configuration is needed on standard GitHub-hosted runners. If your workflow
216- uses a strict network firewall, add `github.com` to the allowed domains:
217-
218- ```yaml
219- network:
220- allowed:
221- - github.com
222- ```
223- -->
0 commit comments