Cowork Skills

ZhangHanDong/makepad-skills: https://github.com/ZhangHanDong/cowork-skills

CLI tool for managing Claude Code skills across 16+ AI coding agents

Crates.io License: MIT Claude Code

Agent Support Status

AgentStatusNotes
Claude CodeFully TestedAll features verified
CursorCommunityContributions welcome
CodexCommunityContributions welcome
GitHub CopilotCommunityContributions welcome
WindsurfCommunityContributions welcome
GooseCommunityContributions welcome
AmpCommunityContributions welcome
RooCommunityContributions welcome
Kiro CLICommunityContributions welcome
Gemini CLICommunityContributions welcome
OpenCodeCommunityContributions welcome
AntigravityCommunityContributions welcome
ClawdbotCommunityContributions welcome
DroidCommunityContributions welcome
KiloCommunityContributions welcome
TraeCommunityContributions welcome

Note: CoWork Skills has been fully tested with Claude Code. Support for other agents is based on documented skill directory conventions. We need community help to test and contribute support for these agents! If you use any of these tools, please help us verify compatibility and submit PRs.

What is CoWork Skills?

CoWork Skills provides a CLI tool (cowork / co) for managing skills across multiple coding agents:

  • Install skills from GitHub repositories to 16+ AI agents
  • Generate skills from source code (Rust, TypeScript, Python)
  • Project-level configuration with Skills.toml
  • Security auditing and checksum verification
  • Search GitHub for skill repositories
  • Manage Claude Code marketplace plugins

Quick Start

Installation

One-line install (recommended):

1curl -sSL https://raw.githubusercontent.com/ZhangHanDong/cowork-skills/main/install.sh | bash

Or install manually:

1# Option 1: From crates.io
2cargo install cowork
3
4# Option 2: From source
5git clone https://github.com/ZhangHanDong/cowork-skills
6cd cowork-skills/cli
7cargo install --path .
8
9# Initialize built-in skills
10cowork init

Install Skills

1# Install from GitHub
2cowork install user/repo
3
4# Install specific skills
5cowork install user/repo -s skill1 -s skill2
6
7# Install to specific agents
8cowork install user/repo -a claude-code -a cursor
9
10# Install as plugin (preserves full repo structure)
11cowork install user/repo --plugin
12
13# Install to project local (.claude/skills/)
14cowork install user/repo --local
15
16# Update to latest version
17cowork install user/repo --update
18
19# List installed repositories
20cowork install --list

CLI Commands

CommandDescription
cowork initInstall built-in skills to ~/.claude/skills/
cowork installInstall skills from GitHub or local path
cowork generateGenerate skills from a GitHub repository
cowork searchSearch GitHub for skill repositories
cowork pluginsManage Claude Code marketplace plugins
cowork configManage project-level skill configuration
cowork listList all available skills
cowork statusShow current configuration
cowork doctorCheck for configuration issues
cowork testGenerate and run trigger tests for skills
cowork auditSecurity audit of installed skills
cowork verifyVerify checksums of installed skills

Use co as a short alias for cowork.

Generate Skills from Source Code

Generate skills from any GitHub repository by parsing source code:

1# Generate Rust skills
2cowork generate tokio-rs/tokio --lang rust
3
4# Generate TypeScript skills
5cowork generate vercel/next.js --lang typescript
6
7# Generate only llms.txt
8cowork generate user/repo --llms-only -o ./output

Supported Languages

LanguageParserExtracts
Rustsynpub fn, struct, enum, trait, impl
TypeScripttree-sitterexport function, class, interface, type
Pythontree-sitterdef, class (excluding _ private items)

Search for Skills

1# Search by keyword
2cowork search tokio
3
4# Search by GitHub topic
5cowork search agent-skill --topic
6
7# Show detailed results
8cowork search rust-skills --verbose

Manage Plugins

1# List marketplace plugins
2cowork plugins list
3
4# Show plugin status
5cowork plugins status
6
7# Uninstall a plugin
8cowork plugins uninstall rust-skills
9
10# Enable/disable plugins
11cowork plugins enable rust-skills
12cowork plugins disable rust-skills

Project Configuration (Skills.toml)

Manage project-level skill configuration with Skills.toml:

1# Initialize config (auto-detects installed plugins/skills)
2cowork config init
3
4# Skip auto-detection
5cowork config init --no-detect
6
7# Show current configuration
8cowork config show
9
10# Add dependencies
11cowork config add rust-skills ZhangHanDong/rust-skills
12cowork config add makepad user/makepad-skills --plugin --local
13
14# Install all dependencies
15cowork config install
16
17# Sync lock file with config
18cowork config sync
19cowork config sync --update  # Also git pull remote repos
20
21# Enable/disable skills or groups
22cowork config enable rust-core
23cowork config disable rust-domains
24
25# Set trigger priority
26cowork config priority dora-router rust-router
27
28# Override specific trigger
29cowork config override "async" rust-router
30
31# Generate dynamic router
32cowork config router
33cowork config router --hooks    # With auto-triggering hooks
34cowork config router --analyze  # Analyze trigger conflicts
35
36# Generate SKILLS.md from config
37cowork config apply

See Skills.toml Configuration Guide for detailed documentation.

Testing Skills

Test that skill triggers are working correctly:

1# Generate trigger test report
2cowork test
3
4# List all triggers with their skills
5cowork test triggers
6
7# Check for trigger conflicts
8cowork test --check-conflicts
9
10# Run actual trigger tests using Claude
11cowork test --run
12
13# Test specific skills
14cowork test --filter "rust-*" --run
15
16# Limit triggers per skill
17cowork test --run -n 5
18
19# Output formats
20cowork test -o triggers.json --format json
21cowork test -o triggers.yaml --format yaml

Built-in Skills

After running cowork init, the following skills are installed globally:

memory-skills

CoALA cognitive architecture memory system with three core capabilities:

  • /remember - Save information to memory (auto-detects global vs project scope)
  • /recall - Search and retrieve information from memory
  • /summarize-session - Summarize current session and save to episodic memory

Memory is organized into:

  • Semantic Memory - Facts, concepts, domain knowledge
  • Episodic Memory - Session summaries, conversation history
  • Procedural Memory - Workflows, patterns, how-to guides

cowork-guide

Complete CLI usage guide triggered when you mention cowork, Skills.toml, or related commands. Provides inline documentation for all CLI features.

cowork-router

Unified router that automatically routes questions to the appropriate installed plugin/skill based on keywords and context.

code-review

Code review assistant triggered by /review-pr or review PR. Features:

  • Fetches PR diff from GitHub API
  • Analyzes code changes for issues
  • Provides structured feedback with severity levels
  • Suggests improvements following best practices

github-generate

Generate skills from GitHub repositories triggered by /github-generate or generate skill from repo. Parses source code and creates skill files with proper triggers.

Search GitHub for skill repositories triggered by /github-search or search for skills. Finds repositories with agent-skill topic or matching keywords.

Commands

1# List available built-in skills
2cowork init --list
3
4# Install specific built-in skills
5cowork init -s memory-skills -s cowork-guide
6
7# Install to project local
8cowork init --local
9
10# Remove specific skills
11cowork init --remove memory-skills

Supported Agents

Install skills to 16+ coding agents:

AgentFlagAgentFlag
Claude Code-a claude-codeAmp-a amp
Cursor-a cursorAntigravity-a antigravity
Codex-a codexClawdbot-a clawdbot
GitHub Copilot-a github-copilotDroid-a droid
Windsurf-a windsurfGemini CLI-a gemini-cli
Goose-a gooseKilo-a kilo
Kiro CLI-a kiro-cliOpenCode-a opencode
Roo-a rooTrae-a trae
1# Install to multiple agents
2cowork install user/repo -a claude-code -a cursor -a windsurf

Security

CoWork provides supply chain security features to protect against malicious skills:

Security Audit

1# Scan all installed skills for security issues
2cowork audit
3
4# Scan specific locations
5cowork audit --global           # Scan ~/.claude/skills/
6cowork audit --project          # Scan .claude/skills/
7cowork audit --plugins          # Scan installed plugins
8
9# Verbose output with details
10cowork audit --verbose
11
12# Save report to file
13cowork audit -o security-report.md --format markdown
14cowork audit -o report.json --format json
15
16# Auto-fix issues where possible
17cowork audit --fix

Detection capabilities:

  • Dangerous patterns (rm -rf, eval(), curl|sh, sudo)
  • Prompt injection attempts
  • Credential leaks (API_KEY, PRIVATE KEY, password)
  • Suspicious system access
  • Risk levels: SAFE, LOW, MEDIUM, HIGH, CRITICAL

Checksum Verification

1# Verify skills against recorded checksums in Skills.lock
2cowork verify
3
4# Update checksums in lockfile
5cowork verify --update
6
7# Verify specific skill
8cowork verify rust-skills
9
10# Verbose output
11cowork verify --verbose

Security Configuration

Add to Skills.toml:

1[security]
2# Trusted authors (skills from these sources are trusted)
3trusted_authors = ["ZhangHanDong", "anthropics"]
4
5# Custom blocked patterns (regex)
6blocked_patterns = ["dangerous-pattern"]
7
8# Paths to skip during scanning (glob patterns)
9# Use for documentation files that describe security patterns
10skip_paths = [
11    "**/docs/**",
12    "**/examples/**",
13    "**/tests/**",
14]
15
16# Trusted marketplace plugins (skip scanning)
17trusted_marketplaces = ["hookify", "rust-skills"]
18
19# Auto-reject high risk skills
20auto_reject_high_risk = false

Storage Locations

LocationPurpose
~/.cowork/repos/Cloned GitHub repositories
~/.claude/skills/Global skills directory
./skills/Project-local skills

Environment Variables

VariableDescription
GITHUB_TOKENRequired for generate/search commands