Skills / AGENTS.md examples / Cursor rules / Claude Code rules

Use skills for task-specific book guidance.

Start with one mini skill for focused work. Use full as reference for deep sessions. Use nano only when context must stay tiny and always-on.

Version Choice

mini is the default for real agent work and usually the best skill body. full is best as skill reference, audit source, or focused-session material. nano is for very small always-on context.

Skills

Prefer skills when the guidance should activate for a specific task: refactoring, review, DDD modeling, reliability work, legacy code, or data-system changes. Every book directory contains a standard SKILL.md entrypoint that points at mini for normal use and keeps full as deeper reference.

refactoring/
  SKILL.md
  refactoring.mini.md
  refactoring.md

Install Existing Skills

Install all book skills with the open Agent Skills CLI:

npx skills add https://github.com/ciembor/agent-rules-books --all

List available skills without installing:

npx skills add https://github.com/ciembor/agent-rules-books --list

Install one selected skill:

npx skills add https://github.com/ciembor/agent-rules-books --skill refactoring

Delivery Patterns

Pattern	Best for	Version
Skill or command	Refactoring, reviews, migrations, reliability work, DDD modeling, legacy changes.	`mini`, optionally backed by `full`
Always-on project rule	Stable defaults that should affect most tasks.	One `mini`, or `nano` if context is tight
Scoped rule	One directory, file type, or subsystem.	`mini` or `nano`
External knowledge system	Large private docs, changing references, or organization-specific knowledge.	Outside this repository

For large private documentation or organization-specific knowledge, use your team's existing documentation or search system outside this repository.

Codex

Use skills for book-specific workflows. Use root AGENTS.md only for stable project-wide defaults.

project/
  AGENTS.md
  .agents/
    skills/
      refactoring-pass/
        SKILL.md

Portable baseline:

cp clean-code/clean-code.mini.md AGENTS.md

Claude Code

Use Claude Code skills for book-specific workflows. Keep CLAUDE.md small and import a common AGENTS.md only when the team wants the same base layer across tools.

project/
  AGENTS.md
  CLAUDE.md
  .claude/
    skills/
      legacy-change/
        SKILL.md

Import the shared root AGENTS.md baseline by adding this line to CLAUDE.md in order to reuse the same project-wide rules in Claude Code.

@AGENTS.md

Cursor

Use Agent Skills for book-specific workflows and .cursor/rules for Cursor-specific always-on, intelligent, manual, or file-scoped prompt rules.

project/
  .agents/
    skills/
      refactoring/
        SKILL.md
  .cursor/
    skills/
      release-it/
        SKILL.md
    rules/
      base.mdc      # Always Apply
      refactor.mdc  # Apply Manually
      ddd.mdc       # Apply Intelligently
      payments.mdc  # Apply to Specific Files

Decision Guide

Need	Use
Task-specific book bias	Create or invoke a skill from the relevant `mini` file.
Steady repo-wide bias	Use one `mini` rule set.
Extremely small always-on baseline	Use `nano`.
Deep audit or derivation	Use `full` as reference.
More than one book	Keep one primary always-on rule and move the rest to skills or on-demand rules.

GitHub Copilot

Use the Markdown content as custom instructions. Keep it short enough to preserve room for repository-specific conventions.