Hermes Core Skills

Your AI coding agent keeps doing dumb things? Here's the fix.

😤 You've been there:

• #1 Agent does /clear and forgets everything — you re-explain from scratch
• #2 Agent sees "fix this bug" and starts changing code — one fix, three new bugs
• #3 Token burns fast — $5 gone before lunch, and you don't know why
• #4 Context fills up and agent starts forgetting what you said 5 minutes ago
• #5 Agent pushes code changes without asking — you only find out when it breaks

This pack of 26 executable skills fixes all of that. Not vague advice. Step-by-step workflows your agent loads and follows.

🎬 Demo

Watch the full 30-second quick install — click to play!

SVG animation — works everywhere, no loading time, scales perfectly.

🚀 Quick Start (30 seconds)

cp -r skills/* ~/.claude/skills/   # Claude Code
cp -r skills/* ~/.cursor/skills/   # Cursor
cp -r skills/* ~/.codex/skills/    # OpenAI Codex
# Hermes Agent: auto-dispatched from ~/.hermes/skills/

Then tell your agent: "Use the systematic-debugging skill to help me with this error."

💡 Why These Skills?

The problem

AI coding agents are powerful, but they have a fatal flaw: they don't know how to work safely yet. Without guidance, an agent will:

• Burn through your token budget in minutes
• Fix one bug and introduce three more
• Push code changes without asking
• Forget context after every /clear
• Loop forever on the same error

The solution

These 25 skills are executable workflows that teach your agent how to behave — systematic debugging, token-aware planning, self-regulation, and safety nets. They're not vague advice; they're step-by-step instructions your agent follows automatically.

🆚 How This Compares

Hermes Core Skills is not a framework you integrate — it's a workflow layer your agent loads. It works alongside any agent and any framework.

💬 Example: How to Use These Skills

Scenario 1: Your agent keeps making bugs worse

You: "Use the systematic-debugging skill on this error: TypeError: Cannot read ..."
Agent: [Loads the 4-stage workflow]
  1. Report — captures the full error context
  2. Context — reads affected files
  3. Hypothesis — identifies root cause
  4. Fix — only applies change after root cause is confirmed
Result: One fix, zero new bugs.

Scenario 2: You're tired of surprise token bills

You: "Follow the token-efficiency skill for this task."
Agent: [Loads token-saving rules]
  - Compresses long context before proceeding
  - Delegates heavy research to sub-agents
  - Avoids wasteful pattern loops
Result: Uses 40-60% fewer tokens.

Scenario 3: Complex project needs multiple features

You: "Run subagent-driven-development to implement the auth system."
Agent: [Splits into parallel sub-agents]
  - Sub-agent 1: Login page
  - Sub-agent 2: JWT middleware
  - Sub-agent 3: Database schema
  - Review pass: merges all, fixes conflicts
Result: Features built in parallel, completed faster.

✨ Highlight Skills

📋 What's Inside

🧠 Agent Core (13 skills)

🤖 AI Agent Ecosystem (6 skills)

🔄 Cross-Session / Planning (3 skills)

🎯 Agent Integration (3 skills)

🏆 Use Cases

📱 Compatible Platforms

🧑‍🎓 For Beginners

New to AI coding agents? Here's everything you need to know.

1. What is an AI coding agent? A tool like Claude Code or Cursor that can write, edit, and debug code for you in your terminal or editor.

2. What's a "skill"? A skill is a Markdown file that teaches your agent how to do something properly — like a recipe for your AI chef.

3. How do I use these skills?

   • Install with one command (cp -r skills/* ~/.claude/skills/)
   • Tell your agent: "Use the [skill name] skill"
   • Your agent follows the instructions automatically

4. Which skill should I start with?

   • First: systematic-debugging — the most useful for daily coding
   • Second: token-efficiency — saves you money immediately
   • Third: checkpoints-and-rewind — safety net, never lose work

Still confused? Open a discussion — we'll help you get started.

🌐 Translations

Help translate this README! Click a badge to contribute:

Currently English only. Translations welcome — submit a discussion or PR!

🗺️ Roadmap

• v1.0.0 — 25 core skills released (Jun 2026)
• v1.1.0 — Skill index (index.json) for agent discovery
• v1.2.0 — Interactive playground on GitHub Pages
• v2.0.0 — Community-contributed skills + skill templates
• Add mappings to common frameworks (pain point categories)

💬 What People Are Saying

"Finally — skills that actually tell the agent what to do instead of just giving it vague instructions." — Early adopter feedback

"The self-regulation-brake-system alone saved me from a $50 runaway agent bill." — Solo developer, Jun 2026

"Drop-in ready and zero config. This is what agent tooling should be." — Open source contributor

Have feedback? Open a discussion and share your experience.

⭐ Star History

📖 Featured In

...

Want to feature this project? Let us know.

🤝 How to Contribute

We'd love your help! Here's how to get started:

Good first issues (no coding needed):

• 📝 Review a skill — Try one and open an issue with your feedback
• 🌐 Translate README — Pick your language and submit a translation
• 🐛 Report a bug — If a skill doesn't work with your agent, let us know
• ✨ Suggest new skills — Open a discussion with your pain point

Code contributions:

1. Browse open issues — look for good first issue labels
2. Fork the repo and create a feature branch
3. Make your changes and submit a PR
4. Your PR will be reviewed within 48 hours

All contributors are expected to follow our Code of Conduct. See the Contributing Guide for details.

📄 Citation

If you use this project in your work:

@software{hermes_core_skills,
  author = {Lam, Chris},
  title = {Hermes Core Skills},
  year = {2026},
  url = {https://github.com/ChrisLamDev/hermes-core-skills},
  note = {25 executable AI agent skills for debugging, planning, token efficiency, and security. MIT licensed.}
}

🛡️ Security

Found a security vulnerability? Please review our Security Policy for responsible disclosure.

📄 License

MIT — Use freely, contribute back when you can.