Agent Skills

Curated local skill registry for Codex, Claude Code, Cursor, and other coding agents.

This repo packages 1,300+ reusable skills with project-local install scripts, manifest support, and a safer workflow for loading only the skills a project actually needs.

Add this to the target project's main AGENTS.md or CLAUDE.md:

Skills: once requirements are clear, load related skills from `/Users/maxpetrusenko/Desktop/Projects/skills/AGENTS.md`.
If requirements are not clear yet, ask the user which skill or domain is needed first.

Why This Repo

• Large local catalog of practical agent skills
• Project-local installs instead of global tool drift
• Manifest-based loading for repeatable setups
• Works across codex, claude, cursor, and .agents
• Built for real coding workflows, not generic prompt dumps

Installation

Clone the registry:

git clone https://github.com/maxpetrusenko/skills.git
cd skills

Quick Start

Install a few skills into a target project:

mkdir -p /absolute/path/to/target-project/.agents

cat > /absolute/path/to/target-project/.agents/skills.task.txt <<'EOF'
# Example task manifest
react-dev
responsive-design
e2e-testing
systematic-debugging
EOF

./scripts/load-skills.sh \
  --project /absolute/path/to/target-project \
  --skills-repo "$(pwd)" \
  --codex-home /absolute/path/to/target-project/.codex \
  --targets codex,agents \
  --manifest /absolute/path/to/target-project/.agents/skills.task.txt \
  --force

After loading, restart the agent session in the target project so it picks up the installed skills.

How To Use It

1. Create a task manifest

Put the exact skills needed for the current project in:

• .agents/skills.must.txt
• .agents/skills.good.txt
• .agents/skills.task.txt

Minimal example:

@domains: frontend, testing, debugging
react-dev
responsive-design
e2e-testing
systematic-debugging
verification-before-completion

2. Load skills into the target project

./scripts/load-skills.sh \
  --project /absolute/path/to/target-project \
  --skills-repo "$(pwd)" \
  --codex-home /absolute/path/to/target-project/.codex \
  --targets codex,claude,cursor,agents \
  --manifest /absolute/path/to/target-project/.agents/skills.task.txt \
  --lockfile /absolute/path/to/target-project/.agents/skills.lock.json \
  --strict-manifest \
  --require-lock \
  --require-domain-coverage \
  --ensure-gitignore \
  --force

3. Clear old installs when the manifest changes

./scripts/clear-skills.sh \
  --project /absolute/path/to/target-project \
  --codex-home /absolute/path/to/target-project/.codex \
  --targets codex,claude,cursor,agents \
  --all \
  --yes

What's Included

• API and backend skills
• Frontend, design, and accessibility skills
• Testing, debugging, and verification skills
• Security, deployment, and CI/CD skills
• Product, planning, and documentation workflows
• Many vendor and integration-specific skills

Every skill lives under skills/ and is usually rooted at:

skills/
└── <skill-name>/
    ├── SKILL.md
    ├── README.md
    ├── references/
    ├── scripts/
    ├── templates/
    └── assets/

Supported Targets

• codex -> <project>/.codex/skills
• claude -> <project>/.claude/skills
• cursor -> <project>/.cursor/skills
• agents -> <project>/.agents/skills

Repo Structure

repo-root/
├── skills/<skill-name>/SKILL.md
├── scripts/load-skills.sh
├── scripts/clear-skills.sh
├── AGENTS.md
└── CLAUDE.md

Supply Chain Notes

• Prefer allowlisted sources before vendoring external skills
• Review SKILL.md, scripts/, and dependency files before loading outside code
• Pin external sources to a commit SHA in the target project lockfile
• Use project-local installs to avoid leaking skills across unrelated repos

Resources

• Main operating guide: AGENTS.md
• Model/runtime notes: CLAUDE.md
• Loader script: scripts/load-skills.sh
• Clear script: scripts/clear-skills.sh

License

No OSS license has been added yet.