LLM Agent Skills

A personal collection of reusable skills for LLM coding agents.

Skills are small, focused instruction packs that teach an agent how to perform a specific workflow consistently. A good skill captures the judgment, conventions, commands, file layout, and verification steps that should be reused across projects.

Goals

• Keep useful agent workflows portable across machines and projects.
• Capture repeatable setup, debugging, writing, and automation patterns.
• Prefer concise instructions over large prompt dumps.
• Make skills easy to inspect, edit, and version.
• Treat skills as living documentation for how I like agents to work.

Repository Layout

skills/
  README.md
  skills/
    example-skill/
      SKILL.md

Each skill should live in its own directory and include a SKILL.md file. Supporting scripts, templates, references, or assets can live next to that file when they make the skill easier to use.

Skill Format

A minimal skill should include:

• A short name and description.
• When the skill should be used.
• The workflow the agent should follow.
• Any commands, scripts, or files the agent should prefer.
• Verification steps, if the skill changes files or system state.

Example

skills/
  macos-setup/
    SKILL.md
    scripts/
      apply-defaults.sh

Writing Principles

• Be specific about triggers.
• Keep instructions operational.
• Prefer checklists and concrete commands.
• Avoid vague personality guidance.
• Include safety notes for destructive or irreversible steps.
• Keep generated artifacts and scratch work out of the repo unless they are part of the skill.

License

Unlicensed for now. Add a license later if these skills become public or shared.