Agent Skills

A collection of opinionated skills for AI coding agents. Built for Claude Code and compatible with OpenClaw.

These are practical, production-tested skills covering game development, software architecture, executive analysis, product illustration, and more. They encode real workflows and domain expertise — not generic advice.

Skills

Skill	Description
concept-art	Iterative concept art generation using ComfyUI (FLUX models, prompt engineering, batch workflows)
corporate-vp	Executive cabinet advisor — analyzes decisions through six C-suite lenses (CEO, COO, CFO, CMO, CPO, CTO)
game-designer	Professional game design — mechanics, systems, GDDs, feature specs, balancing
multi-agent-chat	Coordination protocol for AI agents sharing group chat channels (anti-loop, routing, token efficiency)
product-illustrator	Product renders from CAD files (CadQuery, OpenSCAD) + ComfyUI for ecommerce, manuals, and marketing
software-architecture	Software architecture partner — system design, ADRs, tech specs, code review, security
ue5-character	UE5 character pipeline — modeling, rigging, AnimBP, IK, Control Rig, morph targets
ue5-cinematics	UE5 cinematography — Sequencer, cameras, cutscenes, MRQ rendering
ue5-gamedev	UE5 game development — C++ coding standards, Blueprints, Remote Control API, Python bridge
ue5-level-design	UE5 level design — world building, landscape, lighting, World Partition, streaming
zenstack-v3	ZenStack v3 ORM — schema language, access control, CRUD API, server adapters, migrations

Installation

Claude Code

Clone this repo (or add as a submodule) into your project's skills directory:

# As a git submodule (recommended — stays updated)
git submodule add https://github.com/edhahn/agent-skills.git skills/agent-skills

# Or clone directly
git clone https://github.com/edhahn/agent-skills.git skills/agent-skills

Claude Code discovers skills from skills/ directories automatically. Once installed, skills trigger based on their description field in the YAML frontmatter — no additional configuration needed.

OpenClaw

Same approach — clone or submodule into your workspace's skills/ directory:

git submodule add https://github.com/edhahn/agent-skills.git skills/agent-skills

OpenClaw uses the same skill discovery mechanism as Claude Code.

Cherry-picking individual skills

If you only want specific skills, copy the skill's directory into your skills/ folder:

# Example: just the software-architecture skill
cp -r agent-skills/software-architecture skills/software-architecture

Skills with references/ subdirectories need the entire directory — the main SKILL.md references those files.

Customization

Many skills have sections marked for customization (infrastructure config, team roles, tool references). Look for:

• Environment variable patterns (${COMFYUI_BASE_URL}, UE5_REMOTE_CONTROL_HOST)
• Placeholder tables with generic roles
• Comments suggesting you fill in your own setup

Fork the repo and customize these for your team, or override them in your project's CLAUDE.md.

Skill structure

Each skill follows the Claude Code skill format:

skill-name/
  SKILL.md              # Main skill file with YAML frontmatter
  references/           # Optional deeper reference docs
    topic.md

The SKILL.md frontmatter defines when the skill triggers:

---
name: skill-name
description: >
  When to trigger this skill. Claude Code matches this against
  the current task to decide whether to invoke the skill.
---

Contributing

1. Fork this repo
2. Create a branch for your skill or improvement
3. Follow the structure: skill-name/SKILL.md with YAML frontmatter (name, description)
4. Keep it generic: No personal infrastructure, machine names, agent names, or org-specific details. Use environment variables or placeholder comments for configurable values.
5. Include references: For complex skills, break deep content into references/ files and reference them from the main SKILL.md
6. Write good descriptions: The description field in frontmatter is how Claude Code decides when to trigger the skill. Be specific about trigger conditions.
7. Submit a PR with a brief explanation of what the skill does and why it's useful

What makes a good skill

• Encodes real domain expertise, not generic advice you'd find in a tutorial
• Has a clear trigger condition — when should Claude use this skill vs. not?
• Is actionable — provides workflows, templates, code patterns, and decision frameworks
• Works standalone — doesn't depend on proprietary tools or infrastructure
• Includes reference files for deep dives without bloating the main skill

What to avoid

• Skills that are just documentation dumps (link to docs instead)
• Skills tied to a specific company's infrastructure
• Overly broad skills that try to cover everything ("programming best practices")
• Skills that duplicate Claude's built-in capabilities without adding domain expertise

License

MIT