Helps users discover and install agent skills from the open skills ecosystem (skills.sh). Use when users ask 'how do I do X', 'find a skill for X', 'is there a skill that can...', want to search for tools/templates/workflows, or express interest in extending agent capabilities.
Discover and install agent skills from the skills.sh open ecosystem into Skill Compose.
Activate when users:
find_skills.pySearch the skills.sh ecosystem for skills matching a query.
python scripts/find_skills.py <query> [--limit N]
Example:
python scripts/find_skills.py "react performance"
python scripts/find_skills.py "docker" --limit 5
Output: JSON array of matching skills with name, source (owner/repo), installs count, and url (skills.sh link).
add_skill.pyDownload a skill from GitHub and register it in Skill Compose.
python scripts/add_skill.py <owner/repo@skill-name>
Example:
python scripts/add_skill.py "vercel-labs/agent-skills@vercel-react-best-practices"
What it does:
owner/repo@skill-name identifierskills/<name>/, <name>/, root)skills/ directoryWhen you ask the user a question or present results for them to review, your response MUST end with text only. Do NOT include any tool call (execute_code, bash, etc.) in the same response. The user needs a chance to read and reply. If you combine a question with a tool call, the tool executes immediately without waiting — this breaks the conversation flow.
WRONG (never do this):
"Should I install X?" + [execute_code: install X]
CORRECT (always do this):
Turn 1: "Should I install X?" (text only, no tool calls) Turn 2: User says "yes" Turn 3: [execute_code: install X]
Each step below MUST be a separate conversation turn. Never combine steps.
Identify what domain and specific task the user needs help with.
Run find_skills.py with relevant keywords. Try multiple queries if the first doesn't yield good results. Even if the user names an exact skill, always search first to find the correct source/owner and verify it exists.
Show the user the found skills with:
Ask which skill(s) they want to install. End your response here — no tool calls.
When the user picks a skill, repeat back what you will install and ask for confirmation. This must be a text-only response with no tool calls. Wait for the user to reply.
Only after the user confirms in a separate message, run add_skill.py. Report the result.
| Category | Example Queries | |----------|----------------| | Web Development | react, nextjs, vue, css, tailwind, html | | Testing | testing, jest, playwright, cypress | | DevOps | docker, kubernetes, ci-cd, terraform | | Documentation | docs, readme, markdown, api-docs | | Code Quality | lint, refactor, code-review, typescript | | Design | ui, design, figma, accessibility | | Data & ML | pandas, data-analysis, machine-learning | | Productivity | git, automation, workflow |
vercel-labs/agent-skills, google-labs-code/stitch-skillsskill-creatorSearch for places (restaurants, cafes, etc.) via Google Places API proxy on localhost.
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Start voice calls via the OpenClaw voice-call plugin.
Notion API for creating and managing pages, databases, and blocks.
Gemini CLI for one-shot Q&A, summaries, and generation.
Category:developer