SkillMeat CLI Skill
Natural language interface for managing Claude Code artifacts and environments.
Quick Start
Common Operations
| User Says | What Happens |
|-----------|--------------|
| "Add the PDF skill" | Adds ms-office-suite:pdf to collection |
| "Deploy canvas to this project" | Deploys canvas-design skill to .claude/skills/ |
| "What skills do I have?" | Lists artifacts in current collection |
| "Search for database skills" | Searches all sources for database-related skills |
| "Update all my skills" | Syncs collection with upstream sources |
| "Remove the xlsx skill" | Removes artifact from collection |
For AI Agents
When you identify a capability gap during development:
- Search for relevant artifacts (don't announce this)
- Suggest to user: "This task would benefit from the X skill. Would you like me to add it?"
- Wait for explicit permission before deploying
- Deploy only what was approved
Never auto-deploy artifacts without user permission.
Workflows
Discovery: Finding Artifacts
When user needs a capability or asks what's available:
# Search all sources
skillmeat search "<query>" --type skill
# Search with JSON output (for parsing)
skillmeat search "<query>" --type skill --json
# List what's in collection
skillmeat list --type skill
# Show artifact details
skillmeat show <artifact-name>
Artifact Types: skill, command, agent, mcp, hook
Common Sources:
anthropics/skills/*- Official Anthropic skillsanthropics/example-skills/*- Example/template skills- Community sources (user-configured)
Deployment: Adding Artifacts
When user wants to add or deploy an artifact:
Step 1: Add to Collection
# Add from official source
skillmeat add skill anthropics/skills/canvas-design
# Add specific version
skillmeat add skill anthropics/skills/canvas-design@v1.0.0
# Add from any GitHub repo
skillmeat add skill username/repo/path/to/skill
Step 2: Deploy to Project
# Deploy to current project
skillmeat deploy <artifact-name>
# Deploy to specific project
skillmeat deploy <artifact-name> --project /path/to/project
# Check what's deployed
skillmeat list --project .
Management: Updating & Removing
# Check for updates
skillmeat diff <artifact-name>
# Update specific artifact
skillmeat update <artifact-name>
# Update all artifacts
skillmeat sync --all
# Remove from collection
skillmeat remove <artifact-name>
# Undeploy from project
skillmeat undeploy <artifact-name> --project .
Bundles: Sharing Setups
# Create bundle from current collection
skillmeat bundle create my-setup
# Sign bundle for distribution
skillmeat sign create my-setup.zip
# Import bundle
skillmeat bundle import setup.zip
# Verify bundle signature
skillmeat sign verify setup.zip
claudectl Alias
Power users can use claudectl for simplified commands with smart defaults:
claudectl add pdf # → skillmeat add skill anthropics/skills/pdf
claudectl deploy pdf # → skillmeat deploy pdf --project .
claudectl search database # → skillmeat search database --type skill
claudectl status # → skillmeat list --project . --json
claudectl sync # → skillmeat sync --all
claudectl bundle my-setup # → skillmeat bundle create my-setup
Setup claudectl
Add to ~/.bashrc or ~/.zshrc:
alias claudectl='skillmeat'
Or for smart defaults wrapper, see claudectl-setup.md.
AI Agent Guidelines
When to Suggest Artifacts
DO suggest when:
- User explicitly asks about capabilities
- Clear capability gap identified (e.g., "process this PDF" but no PDF skill)
- User asks for help with a task that has a well-known skill
DON'T suggest when:
- Task can be completed without additional skills
- User is in the middle of focused work
- Suggestion would be redundant (skill already deployed)
Suggestion Format
I notice this task involves [capability]. The [artifact-name] skill
could help with this. Would you like me to add it to the project?
It provides: [brief description]
Source: [source path]
Permission Protocol
- Always ask before deploying
- Show what will change (files to be created)
- Confirm success after deployment
- Suggest related artifacts only if relevant
Command Reference
For complete command documentation, see command-quick-reference.md.
Most Used Commands
| Command | Purpose |
|---------|---------|
| skillmeat search <query> | Find artifacts |
| skillmeat add <type> <source> | Add to collection |
| skillmeat deploy <name> | Deploy to project |
| skillmeat list | List in collection |
| skillmeat list --project . | List deployed in project |
| skillmeat show <name> | Show artifact details |
| skillmeat sync | Sync with upstream |
| skillmeat remove <name> | Remove from collection |
Artifact Resolution
The skill resolves fuzzy names to full identifiers:
| User Says | Resolves To |
|-----------|-------------|
| "pdf" | ms-office-suite:pdf or example-skills:pdf |
| "canvas" | canvas-design |
| "xlsx" | ms-office-suite:xlsx |
| "docx" | ms-office-suite:docx |
When ambiguous, present options and ask user to choose.
Project Context Analysis
When recommending artifacts, analyze project context:
| Project Indicator | Recommended Artifacts |
|-------------------|----------------------|
| package.json with React | frontend-design, webapp-testing |
| pyproject.toml | Python-related skills |
| FastAPI imports | openapi-expert, backend skills |
| .claude/ directory | Check what's already deployed |
| tests/ directory | Testing-related skills |
See analyze-project.js for analysis script.
Error Handling
Common Issues
| Error | Solution |
|-------|----------|
| "Artifact not found" | Check spelling, try search first |
| "Already in collection" | Use deploy to deploy existing artifact |
| "Permission denied" | Check directory permissions |
| "Rate limited" | Set GitHub token: skillmeat config set github-token <token> |
Getting Help
skillmeat --help # General help
skillmeat <command> --help # Command-specific help
skillmeat web doctor # Diagnose environment issues
Related Skills
- skill-builder: Create new skills
- skill-creator: Design skill workflows
- chrome-devtools: Browser automation (example of CLI wrapper skill)
微信扫一扫