Skill Discovery#

Automatically search ClawHub for skills that match the user's needs and offer to install them.

When to Activate#

• No local skill matches the user's request
• User explicitly asks to find or install a skill
• Agent determines a specialized skill would handle the task better than general-purpose reasoning

Workflow#

1. Identify the Need#

Extract 1-3 concise search keywords from the user's request. Focus on the core capability, not the full sentence.

Examples:

• "帮我管理 Docker 容器" → docker
• "I need to process CSV files" → csv
• "能不能帮我发邮件" → email

2. Search ClawHub#

bash

openclaw skills search "<keywords>"

If no results, try alternative keywords or broader terms.

3. Present Results#

Show the user what was found in a concise list:

• Skill name / slug
• Brief description
• Relevance to their request

If nothing relevant is found, tell the user honestly and suggest they could create one or check clawhub.ai manually.

4. Install (with user confirmation)#

Always ask the user before installing. Never auto-install.

bash

openclaw skills install <slug>

After install, inform the user:

• The skill is installed
• They need to start a new session (or the agent needs a restart) for the skill to take effect
• Alternatively, they can read the new SKILL.md immediately to use it in the current session

5. Immediate Use (optional)#

If the user wants to use the skill right away in the current session:

1. Read the newly installed skill's SKILL.md
2. Follow its instructions to handle the original request

Guidelines#

• Don't over-search: Only trigger when there's a genuine gap. If the task can be done with existing tools (exec, read, web_fetch, etc.), just do it.
• Be specific: Use precise keywords. "weather" not "get the current weather forecast for my city".
• Respect the user: Always confirm before installing. Show what you found, let them decide.
• One at a time: Search and install one skill per request. Don't bulk-install.