agentleFS
ZhangHanDongZhangHanDong/cowork-skills/skills/cowork-guide/SKILL.md

cowork-guide

Skill46 starsChanged 9 months agoMITView on GitHub

CRITICAL: Comprehensive guide for CoWork Skills CLI tool. Triggers on: cowork, Skills.toml, skill management, plugin configuration, cowork init, cowork install, cowork config, cowork generate, cowork audit, cowork verify, cowork test

  • Pipes a download into a shell
  • Reads credentials
  • Deletes or force-pushes
  • Installs packages

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 2 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Search GitHub for skill repositories

    line 17
  2. Distinctive:

    Search GitHub for skill repositories.

    line 504
  3. Very specific:

    Verify checksums of installed skills against Skills.lock.

    line 636
  4. Very specific:

    Check if skill exists: cowork list

    line 716
  5. Very specific:

    Verify enabled status: cowork config show

    line 717
  6. Very specific:

    Ensure triggers match: cowork test triggers

    line 718
  7. Very specific:

    Check GitHub token: echo $GITHUB_TOKEN

    line 721
  8. Very specific:

    Verify repo exists: gh repo view user/repo

    line 722
  9. Very specific:

    Check for conflicts: cowork test --check-conflicts

    line 728

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

CoWork Guide

Complete reference for the CoWork Skills CLI tool

Overview

CoWork Skills (cowork / co) is a CLI tool for managing Claude Code skills and plugins across 16+ AI coding agents:

  • Install skills from GitHub repositories
  • Generate skills from source code (Rust, TypeScript, Python)
  • Manage project-level skill configuration via Skills.toml
  • Security auditing and checksum verification
  • Search GitHub for skill repositories

Installation

First, check if cowork is installed:

cowork --version

If not installed, install via one of these methods:

# Option 1: From crates.io (recommended)
cargo install cowork

# Option 2: From source
git clone https://github.com/ZhangHanDong/cowork-skills
cd cowork-skills/cli
cargo install --path .

Then initialize built-in skills:

cowork init

Note: If you installed cowork-skills via npx @anthropic-ai/claude-code add, you still need to install the cowork CLI separately using the commands above.

Command Alias

cowork = co   # 简写别名

Commands by Category

🚀 Quick Start (初始化)

CommandDescription
co init安装内置 skills 到全局/项目
co init --list列出可用的内置 skills
co init --local安装到项目 .claude/skills/
co init -r <name>删除指定 skill

📦 Install & Uninstall (安装)

CommandDescription
co install user/repo从 GitHub 安装
co install user/repo -s skill1安装指定 skills
co install user/repo --plugin作为插件安装
co install user/repo -l安装到项目本地
co install user/repo --update更新到最新版本
co install --list列出已安装仓库
co install --uninstall repo卸载仓库

⚙️ Configuration (配置)

CommandDescription
co config init创建 Skills.toml
co config show显示当前配置
co config add name repo添加依赖
co config add name path --dev添加开发链接
co config install安装 Skills.toml 中的依赖
co config sync同步锁文件状态
co config enable/disable启用/禁用 skills
co config priority设置触发优先级
co config router生成动态路由
co config router --analyzeAI 分析生成智能路由

🔍 Discovery (发现)

CommandDescription
co list列出所有可用 skills
co search query搜索 GitHub 仓库
co search topic --topic按 topic 搜索
co generate user/repo从源码生成 skills

🔌 Plugin Management (插件)

CommandDescription
co plugins list列出已安装插件
co plugins status插件系统状态
co plugins enable/disable启用/禁用插件
co plugins uninstall卸载插件

🛡️ Security & Testing (安全)

CommandDescription
co audit安全审计
co verify校验 checksums
co test生成触发测试
co test --run运行实际测试
co test --check-conflicts检查触发冲突

📊 Status & Debug (状态)

CommandDescription
co status显示当前状态
co doctor检查配置问题

Workflow 工作流

┌─────────────────────────────────────────────────────────────────┐
│                        CoWork 工作流                              │
└─────────────────────────────────────────────────────────────────┘

1. 初始化
   co init                          # 安装内置 skills

2. 安装外部 skills
   co install user/repo             # 从 GitHub 安装
   co install user/repo --plugin    # 作为插件安装

3. 项目配置 (可选)
   co config init                   # 创建 Skills.toml
   co config add rust ZhangHanDong/rust-skills
   co config install                # 安装依赖

4. 生成路由
   co config router --analyze       # AI 分析生成智能路由

5. 测试 & 验证
   co test --check-conflicts        # 检查触发冲突
   co audit                         # 安全审计

6. 日常使用
   co list                          # 查看所有 skills
   co status                        # 查看状态
   co doctor                        # 诊断问题

cowork init

Initialize built-in skills to ~/.claude/skills/.

# Install all built-in skills
cowork init

# List available built-in skills
cowork init --list

# Install specific skills only
cowork init -s memory-skills -s cowork-guide

# Force overwrite existing
cowork init --force

# Install to project local (.claude/skills/)
cowork init --local

# Remove specific skills
cowork init --remove memory-skills

Built-in Skills:

  • memory-skills - CoALA-based memory system (remember, recall, reflect)
  • cowork-guide - Complete CoWork CLI usage guide
  • cowork-router - Unified router for installed plugins/skills
  • code-review - Code review assistant with best practices
  • github-generate - Generate skills from GitHub repositories
  • github-search - Search GitHub for skill repositories

cowork install

Install skills from GitHub or local repositories.

Basic Usage

# Install from GitHub (user/repo format)
cowork install ZhangHanDong/rust-skills

# Install from full URL
cowork install https://github.com/user/repo

# Install current project to global
cowork install

# List installed repositories
cowork install --list

Installation Options

# Install specific skills only
cowork install user/repo -s skill1 -s skill2

# Install to specific agents
cowork install user/repo -a claude-code -a cursor

# Install as plugin (preserves full repo structure)
cowork install user/repo --plugin

# Install to project local (.claude/skills/)
cowork install user/repo --local

# Force reinstall
cowork install user/repo --reinstall

# Update to latest version
cowork install user/repo --update

# Include additional directories
cowork install user/repo --include-dir docs --include-dir examples

# Copy files instead of symlinks
cowork install user/repo --no-symlink

# Use npx add-skill backend
cowork install user/repo --use-add-skill

# Skip confirmation prompts
cowork install user/repo -y

Uninstall

cowork install --uninstall repo-name

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

cowork config

Manage project-level skill configuration via Skills.toml.

Initialize Configuration

# Initialize with auto-detection of installed plugins/skills
cowork config init

# Skip auto-detection
cowork config init --no-detect

# Force overwrite existing
cowork config init --force

Auto-detection scans:

  • ~/.claude/ for global plugins
  • ~/.claude/skills/ for global skills
  • .claude/ for project plugins
  • .claude/skills/ for project skills

View Configuration

# Show current config
cowork config show

# List available skill groups
cowork config groups

Add Dependencies

# Add from GitHub (global)
cowork config add rust-skills ZhangHanDong/rust-skills

# Add to project local
cowork config add my-skills user/skills --local

# Add specific skills
cowork config add tokio user/tokio-skills -s tokio-runtime -s tokio-sync

# Add as plugin
cowork config add makepad user/makepad-skills --plugin

# Add as disabled
cowork config add old-lib user/old --disabled

# Add with git ref
cowork config add pinned user/repo --ref v1.0.0

# Add development link (symlink for testing)
cowork config add dora-dev /path/to/dora-skills --dev

Install Dependencies

# Install all dependencies from Skills.toml
cowork config install

Sync Configuration

# Sync enabled/disabled status with lock file
cowork config sync

# Also update remote repos (git pull)
cowork config sync --update

Enable/Disable

# Enable skill groups
cowork config enable rust-core makepad

# Enable individual skills
cowork config enable memory-filesystem

# Disable skills or groups
cowork config disable rust-domains

Trigger Configuration

# Set priority order
cowork config priority dora-router rust-router makepad-router

# Override specific trigger
cowork config override "async" rust-router
cowork config override "widget" makepad-router

Generate Output

# Generate SKILLS.md from config
cowork config apply

# Generate to custom path
cowork config apply -o ./docs/SKILLS.md

# Generate dynamic router based on installed plugins
cowork config router

# Generate router with hooks for auto-triggering
cowork config router --hooks

# Analyze trigger conflicts
cowork config router --analyze

# Skip cache
cowork config router --no-cache

Skills.toml

Project-level configuration file stored at .cowork/Skills.toml.

Basic Structure

[project]
name = "my-project"
description = "Project description"

[skills.global]
enabled = ["memory-filesystem"]
disabled = []

[skills.install]
rust-skills = "ZhangHanDong/rust-skills"

[skills.groups]
enabled = ["rust-core"]
disabled = []

[triggers]
priority = ["dora-router", "rust-router"]

[triggers.overrides]
"async" = "rust-router"

Dependency Forms

Simple form:

[skills.install]
rust-skills = "ZhangHanDong/rust-skills"

Detailed form:

[skills.install]
tokio = { repo = "user/tokio-skills", skills = ["tokio-runtime", "tokio-sync"] }
local = { path = "../my-local-skills" }
pinned = { repo = "user/repo", ref = "v1.0.0" }
makepad = { repo = "user/makepad-skills", plugin = true }
my-project = { repo = "user/skills", local = true }
old-lib = { repo = "user/old", enabled = false }

Dev Links

Development symlinks for testing local skills:

[skills.dev]
my-skill = "/path/to/my-skill-project"
dora-dev = { path = "/path/to/dora-skills" }
dora-plugin = { path = "/path/to/dora-skills", plugin = true }

Skill Groups

GroupSkillsDescription
rust-core8Basic Rust (ownership, concurrency, error handling)
rust-patterns7Design patterns (domain modeling, performance)
rust-domains7Domain-specific (web, CLI, fintech, embedded)
makepad11Makepad UI framework
dora8Dora-rs robotics framework
dora-hubs9Dora hub integrations

cowork generate

Generate skills from source code of any GitHub repository.

# Generate from GitHub repo
cowork generate user/repo

# Specify language(s)
cowork generate tokio-rs/tokio --lang rust
cowork generate vercel/next.js --lang typescript

# Only generate llms.txt
cowork generate user/repo --llms-only -o ./output

# Generate from existing llms.txt
cowork generate --from-llms ./llms.txt

# Specify git ref
cowork generate user/repo --ref v1.0.0

Supported Languages

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

Output

  1. llms.txt - API documentation following llms.txt specification
  2. SKILL.md - Generated skill files with triggers and references

cowork search

Search GitHub for skill repositories.

# Search by keyword
cowork search tokio

# Search by GitHub topic
cowork search agent-skill --topic

# Limit results
cowork search rust-skills --limit 5

# Show detailed results
cowork search rust-skills --verbose

cowork plugins

Manage Claude Code marketplace plugins.

# List marketplace plugins
cowork plugins list

# Show plugin status
cowork plugins status

# Uninstall a plugin
cowork plugins uninstall rust-skills

# Enable/disable plugins
cowork plugins enable rust-skills
cowork plugins disable rust-skills

# List marketplaces
cowork plugins marketplaces

# Remove a marketplace
cowork plugins remove-marketplace rust-skills

cowork test

Generate and run trigger tests for skills.

Generate Test Reports

# Generate trigger test report
cowork test

# List all triggers with their skills
cowork test triggers

# Check for trigger conflicts
cowork test --check-conflicts

# Output formats
cowork test -o triggers.json --format json
cowork test -o triggers.yaml --format yaml
cowork test -o triggers.md --format markdown

Run Actual Tests

# Run trigger tests using Claude
cowork test --run

# Test specific skills by pattern
cowork test --filter "rust-*" --run

# Limit triggers per skill
cowork test --run -n 5

# Scan specific locations
cowork test --global          # ~/.claude/skills/
cowork test --path ./skills   # Custom directory
cowork test --plugins         # Installed plugins
cowork test --all             # Project + global + plugins

cowork audit

Security audit of installed skills.

Basic Usage

# Scan all installed skills
cowork audit

# Verbose output with details
cowork audit --verbose

# Auto-fix issues where possible
cowork audit --fix

Scan Locations

cowork audit --global    # Scan ~/.claude/skills/
cowork audit --project   # Scan .claude/skills/
cowork audit --plugins   # Scan installed plugins

Output Formats

cowork audit -o report.txt --format text
cowork audit -o report.json --format json
cowork audit -o report.md --format markdown

Detection Capabilities

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

cowork verify

Verify checksums of installed skills against Skills.lock.

# Verify all skills
cowork verify

# Verify specific skill
cowork verify rust-skills

# Update checksums in lockfile
cowork verify --update

# Verbose output
cowork verify --verbose

Security Configuration

Add to Skills.toml:

[security]
# Trusted authors
trusted_authors = ["ZhangHanDong", "anthropics"]

# Custom blocked patterns (regex)
blocked_patterns = ["dangerous-pattern"]

# Paths to skip during scanning (glob patterns)
skip_paths = [
    "**/docs/**",
    "**/examples/**",
    "**/tests/**",
]

# Trusted marketplace plugins (skip scanning)
trusted_marketplaces = ["hookify", "rust-skills"]

# Auto-reject high risk skills
auto_reject_high_risk = false

Storage Locations

LocationPurpose
~/.cowork/repos/Cloned GitHub repositories
~/.claude/skills/Global skills directory
~/.claude/<plugin>/Global plugins
.claude/skills/Project-local skills
.claude/<plugin>/Project-local plugins
.cowork/Skills.tomlProject configuration
.cowork/Skills.lockInstalled packages lock

Environment Variables

VariableDescription
GITHUB_TOKENRequired for generate/search commands

Troubleshooting

Check Configuration

cowork doctor
cowork status

Common Issues

Skills not loading:

  1. Check if skill exists: cowork list
  2. Verify enabled status: cowork config show
  3. Ensure triggers match: cowork test triggers

Installation fails:

  1. Check GitHub token: echo $GITHUB_TOKEN
  2. Verify repo exists: gh repo view user/repo
  3. Try with verbose: cowork install user/repo --reinstall

Plugin conflicts:

  1. Set trigger priority: cowork config priority router1 router2
  2. Override specific triggers: cowork config override "keyword" skill-name
  3. Check for conflicts: cowork test --check-conflicts

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.