使用说明
Awesome Agent Tools 🤖✨
The hybrid marketplace for AI agent development: original skills, curated tools, and cross-platform support.
For developers building with AI agents | Original Skills + Curated Tools | Cross-Platform
Supports: Claude Code (⭐ Recommended) | Codex | Cursor | GitHub Copilot | Antigravity | Devin Desktop | Other Agent Skills clients
Why This Collection?
Unlike other lists, we offer:
- 🔀 Hybrid Marketplace: Original skills we built + carefully curated third-party tools
- 👨💻 For Developers: Built by developers, for developers building with AI agents
- 🌐 Cross-Platform: Skills use the open Agent Skills format, so they work in Claude Code, Codex, Cursor, GitHub Copilot, Antigravity, and other compatible agents
- ⚡ Multi-Type: Skills + MCP Servers + Agents + Plugins (not just one category)
- ✨ Quality-First: Every tool is vetted and tested, not a link dump
- 🎯 Agent Development: Special focus on agent workflows, parallel execution, and automation
✨ Features
- 🎯 Curated Collection - Only high-quality, well-tested tools and skills
- 🔧 Cross-Platform - Skills work in any Agent Skills-compatible agent: Claude Code, Codex, Cursor, GitHub Copilot, Antigravity, and more
- 📦 Easy Installation - Simple one-command setup for most tools
- 🌍 Community-Driven - Open source and welcoming contributions
- 🚀 MCP-Powered - Leverages the Model Context Protocol standard
- 📚 Well-Documented - Clear instructions and examples for everything
Quick Install
Claude Code (⭐ Recommended)
Install everything (skills, the tech-scout agent, and the handoff command) as one plugin:
# In a Claude Code session
/plugin marketplace add Awakehsh/awesome-agent-tools
/plugin install awesome-agent-tools@awesome-agent-tools
# Or from your shell
claude plugin marketplace add Awakehsh/awesome-agent-tools
claude plugin install awesome-agent-tools@awesome-agent-tools
Plugin components are namespaced, for example /awesome-agent-tools:codex-cli and /awesome-agent-tools:handoff.
Any Agent Skills-Compatible Agent
Install a single skill with the skills CLI:
npx skills add Awakehsh/awesome-agent-tools --skill learn
# --list shows available skills, -g installs for your user, -a picks the agent (e.g. -a codex)
Manual Copy
git clone https://github.com/Awakehsh/awesome-agent-tools
mkdir -p ~/.claude/skills ~/.agents/skills
cp -R awesome-agent-tools/skills/learn ~/.claude/skills/ # Claude Code
cp -R awesome-agent-tools/skills/learn ~/.agents/skills/ # Codex
In Codex you can also run $skill-installer https://github.com/Awakehsh/awesome-agent-tools/tree/main/skills/learn inside a session (not your shell).
Skill folders for other agents: Platform Support. Detailed setup: Cross-Platform Guide
🤖 Agent Development Highlights
Special features for building AI agent systems:
| Feature | Description | Platform |
|---|---|---|
| Custom Sub-Agents | Independent contexts for specialized tasks (guide) | Claude Code |
| Parallel Execution | Run multiple agents simultaneously (examples) | Claude Code |
| Agent Skills | Reusable skill bundles for agents | Any Agent Skills-compatible agent |
| MCP Integration | Connect agents to external tools & data sources | All platforms |
| Workflow Automation | Slash commands, hooks, and agent orchestration | Claude Code |
New to agents? Check out our Agent Development Guide for workflows, patterns, and best practices.
📦 What's Included
🎯 Editor's Picks (Quick Wins)
Start here if you're new - these tools deliver immediate value:
| Tool | Type | What It Does | Platform |
|---|---|---|---|
| ralph-loop | Official Plugin | Overnight automated development with iterative loops | Claude Code |
| learn | Original Skill | Auto-capture insights to project memory (CLAUDE.md, AGENTS.md) | All |
| context-engineering-kit | Community Tool | Advanced context engineering, 8-21% quality improvement | All |
| codex-cli | Original Skill | Call OpenAI Codex from Claude Code for plan review | Claude Code |
| awesome-mcp-servers | Curated List | Widely used community collection of MCP servers | All |
| superpowers | Skill Framework | Complete AI agent development workflow (TDD, code review, debugging) | Claude Code, Codex, Cursor, and more |
| ECC (formerly everything-claude-code) | Config Collection | Battle-tested agents, skills, hooks, and rules from a hackathon winner | Claude Code, Codex, and more |
🛠️ Original Tools
Open-source contributions from this repository:
Skills (3)
context-architect
Context engineering tool for building AI agent systems — design system prompts, tool sets, examples, and runtime context management.
- Purpose: Transform requirements into production-ready Context Engineering Packages through guided workflows
- Philosophy: Context-first design — manage the entire information state (prompts, tools, examples, retrieval, compaction), not just instructions
- Use Cases:
- Design multi-agent orchestration systems with context isolation and sub-agent summary constraints
- Plan token budgets, tool sets, and few-shot examples with engineering rigor
- Manage long-horizon agent sessions with compaction, note-taking, and Just-in-Time retrieval strategies
- Key Benefit: Goes beyond prompt engineering to full context engineering — based on Anthropic's context engineering research
- Install:
npx skills add Awakehsh/awesome-agent-tools --skill context-architect(or the Claude Code plugin)
codex-cli
Call OpenAI Codex CLI from Claude Code for plan review and code analysis.
- Use Cases: Plan review, code analysis, technical discussions, multi-round iteration
- Install:
npx skills add Awakehsh/awesome-agent-tools --skill codex-cli(or the Claude Code plugin)
learn
Universal learning system that extracts insights from conversations and updates project memory files.
- Use Cases: Capture learnings, auto-route to docs (CLAUDE.md, AGENTS.md), cross-platform sync
- Install:
npx skills add Awakehsh/awesome-agent-tools --skill learn(or the Claude Code plugin)
Slash Commands (1)
handoff
Generate a structured handoff for the next coding agent (Codex / Claude / Cursor) picking up your task.
- Use Cases: Cross-agent work transfer, mid-task pause/resume, switching tools mid-feature
- Output: 8-section paste-ready brief — goal, state, decisions, failed approaches, evidence, open questions, next steps, relevant files
- Install: Included in the Claude Code plugin as
/awesome-agent-tools:handoff, ormkdir -p ~/.claude/commands && cp commands/handoff.md ~/.claude/commands/
Agents (1)
tech-scout
Technical research specialist. Searches industry solutions, design patterns, and bug fixes from GitHub and web sources.
- Purpose: Parallel research during development workflows
- Philosophy: Don't waste time reinventing - learn from industry solutions first
- Use Cases:
- Feature planning - Search architecture patterns and design approaches
- Bug debugging - Find solutions for errors and common problems
- Technology comparison - Research pros/cons of libraries and frameworks
- Key Benefit: Runs independently without blocking main workflow, returns structured reports with comparisons and sources
- Install: Included in the Claude Code plugin, or
mkdir -p ~/.claude/agents && cp agents/tech-scout.md ~/.claude/agents/
Learn more → | All Agents Documentation →
MCP Servers (Coming Soon)
- Custom integrations
- Specialized tooling
- Watch this space!
🌟 Curated Recommendations
Carefully vetted third-party tools organized by category:
See the full curated list: AWESOME.md
Categories:
- 🎯 Skills - Official Anthropic skills, collections, and installers
- 🧩 Plugins & Marketplaces - Claude Code marketplaces and notable plugins
- 🔧 MCP Servers - Reference, open-source, and hosted servers
- 🧰 Agent Runtimes, Security, and Evaluation - Orchestrators, memory, scanners, and benchmarks
- 🛠️ Developer Resources - SDKs, frameworks, and community directories
📚 Guides & Documentation
- Agent Development Guide - Workflows, parallel execution, best practices
- Cross-Platform Setup - Skill folders and install steps for each platform
- Build Your First Skill - Step-by-step
SKILL.mdtutorial - Create an MCP Server - Checklist for an MCP server that works across clients
- Contributing Guide - How to add your tools, inclusion criteria, and repository standards
Use Cases
| Category | Examples |
|---|---|
| Development | Code generation, refactoring, testing, debugging |
| Content | Documentation, translation, technical writing |
| Research | Data analysis, summarization, literature review |
| Automation | CI/CD, task automation, monitoring |
Platform Support
Skills in this repository follow the open Agent Skills standard: a folder with a SKILL.md file. These agents read SKILL.md natively:
| Platform | Project skills | User skills |
|---|---|---|
| Claude Code ⭐ | .claude/skills/ |
~/.claude/skills/ |
| Codex | .agents/skills/ |
~/.agents/skills/ |
| Cursor | .agents/skills/ or .cursor/skills/ |
~/.agents/skills/ or ~/.cursor/skills/ |
| GitHub Copilot | .agents/skills/ or .github/skills/ |
~/.agents/skills/ or ~/.copilot/skills/ |
| Antigravity | .agents/skills/ |
~/.gemini/config/skills/ (CLI: ~/.gemini/antigravity-cli/skills/) |
| Devin Desktop (formerly Windsurf) | .agents/skills/ or .devin/skills/ |
~/.agents/skills/ or ~/.config/devin/skills/ |
Every agent above except Claude Code reads project skills from .agents/skills/. Per-platform details and docs links: Cross-Platform Guide. More compatible agents: Agent Skills client showcase.
Why we recommend Claude Code: this repository ships a Claude Code plugin, so one install brings the skills, the tech-scout agent, and the handoff command. Other agents use the skills.
Claude Code Tips
Claude Code has powerful built-in features you might not know:
| Feature | Description |
|---|---|
/skills |
List the skills available in your session and manage their visibility |
/agents |
Create custom sub-agents with independent contexts (learn more) |
CLAUDE.md |
Project memory that prevents repeated mistakes |
| Slash Commands | Store workflows in .claude/commands/ for automation |
| Parallel Agents | Run multiple agents simultaneously for independent tasks (learn more) |
| Plan Mode | Read-only codebase analysis before making changes |
Pro tip from Anthropic team: Add mistakes to CLAUDE.md so Claude learns and doesn't repeat them.
Learn more: Official Docs | Creator's Workflow
Codex Tips
Codex skill loading is folder-based. A skill lives at ~/.agents/skills/<skill-name>/SKILL.md (user) or .agents/skills/<skill-name>/SKILL.md (repository).
| Tip | Description |
|---|---|
$skill-installer ... |
Run this in a Codex session (not your shell) with a skill name or a GitHub URL to a skill folder |
| Verify install | Run /skills in Codex, or type $ to mention a skill |
| Create a new skill | Make ~/.agents/skills/<skill-name>/SKILL.md (YAML frontmatter + Markdown) |
| Reload skills | Codex detects new and changed skills automatically; restart Codex if one doesn't appear |
Resources
Official Documentation
Community Resources
- Awesome MCP Servers - Curated list of MCP servers
- Awesome Claude - Comprehensive directory
- MCP Servers Repository
- MCP Registry - Official registry of publicly available MCP servers (preview)
Contributing
We welcome contributions! Here's how you can help:
- Add Your Own Skill - Share your custom-built skills
- Recommend Tools - Suggest great third-party skills and servers
- Improve Documentation - Fix typos, add examples
- Report Issues - Found a bug? Let us know!
See CONTRIBUTING.md for detailed guidelines.
License
MIT License. See LICENSE for details.
- Our Content: The MIT License covers this repository's original content and documentation
- Third-Party Content: Tools listed in AWESOME.md keep their own licenses; this repository is an index and claims no ownership of them
Made with ❤️ by the community
