Skills
Skills are modular capabilities that extend OpenClaw's functionality. They provide specialized tools for specific domains.
What Are Skills
A skill is a self-contained package that includes:
- Domain knowledge and workflows
- Tool integrations
- Bundled resources (scripts, references, assets)
Skills are stored in:
/app/skills/: Built-in skills (read-only)~/.openclaw/workspace/skills/: Custom skills (user-defined)
Available Skills
github
Interact with GitHub using the gh CLI.
- Clone repositories
- Create issues and PRs
- Check CI runs
- Query repository data
weather
Get current weather and forecasts.
- No API key required
- Location-based queries
- Forecast data
healthcheck
Host security hardening and system checks.
- Security audits
- Firewall/SSH hardening
- Update management
- Risk posture review
cron
Scheduled jobs and reminders.
- Recurring tasks (daily/weekly reports)
- One-shot reminders
- Background agent execution
web_search / web_fetch
Internet research capabilities.
- Brave Search API integration
- URL content extraction
- Markdown/text output
browser
Web automation via Chromium.
- Page navigation and interaction
- Screenshots
- PDF generation
message
Cross-platform messaging.
- Send messages to any configured channel
- Broadcasts, reactions, polls
- Channel management
sessions_spawn
Background agent tasks.
- Spawn isolated sub-agents
- Long-running operations
- Parallel task execution
Using Skills
Automatic Invocation
The model automatically selects appropriate skills based on context. For example:
- "Check the weather" → triggers
weatherskill - "Clone my repo" → triggers
githubskill - "Remind me tomorrow" → triggers
cronskill
Manual Reference
You can explicitly request a skill:
- "Use the github skill to list my repos"
- "Check healthcheck for security issues"
Creating Custom Skills
Structure
skill-name/
├── SKILL.md # Required: metadata + instructions
├── scripts/ # Optional: executable code
├── references/ # Optional: documentation
└── assets/ # Optional: templates, images
SKILL.md Format
---
name: skill-name
description: "What this skill does. Use when..."
tools: Read, Write, Exec
---
# Instructions
...
Progressive Disclosure
Skills use three loading levels:
- Metadata (
name+description): Always in context - SKILL.md body: Loaded when skill triggers
- References/Scripts: Loaded on demand
Keep SKILL.md under 500 lines. Move detailed docs to references/.
Skill Development
- Understand the use case with concrete examples
- Plan reusable resources (scripts, references)
- Initialize with proper structure
- Implement SKILL.md and resources
- Test thoroughly before use
- Package for distribution
Related
- Skill Creator Guide
- Personas: Custom agent behaviors