Skills first
Use mini as the default skill body for reviews, refactors, architecture passes, and domain work. Use full for deep, on-demand skills.
AGENTS.md rules / skills / Cursor rules / Claude Code memory
AI agent rules / skills distilled from software engineering books.
Markdown rule sets for coding agents. Use mini or full as focused skills when a task needs a book's decision pressure, and keep nano for tiny always-on context.
VISIT REPOSITORY
git clone https://github.com/ciembor/agent-rules-books.git
cd agent-rules-books
# Recommended default
cp clean-code/clean-code.mini.md AGENTS.md
# Tight context fallback
cp refactoring/refactoring.nano.md AGENTS.md
# Deep reference session
open domain-driven-design/domain-driven-design.mdTool agnostic
Plain Markdown that can become Codex skills, Claude Code skills, AGENTS.md, CLAUDE.md, Cursor rules, memories, or custom instructions.
Three sizes
mini for focused skills and normal agent work, nano for small always-on context, full for reference skills and audits.
Rules / Skills
Choose one primary rule set or skill for the task. Prefer skills for mini and full; reserve always-on rules for stable project defaults.
| Rule / skill | Best for | Files |
|---|---|---|
| A Philosophy of Software Design | API design, module depth, complexity reduction. | mini / nano / full |
| Clean Architecture | Boundaries, dependency direction, framework isolation. | mini / nano / full |
| Clean Code | Everyday readability, naming, functions, tests. | mini / nano / full |
| Code Complete | Construction discipline, defensive programming, implementation quality. | mini / nano / full |
| Designing Data-Intensive Applications | Reliability, scalability, consistency, schemas, streams. | mini / nano / full |
| Domain-Driven Design | Domain modeling, bounded contexts, ubiquitous language. | mini / nano / full |
| Domain-Driven Design Distilled | Lightweight DDD, subdomains, context maps. | mini / nano / full |
| Implementing Domain-Driven Design | Aggregates, events, contexts, application architecture. | mini / nano / full |
| Patterns of Enterprise Application Architecture | Layers, service layer, data mapper, repository, unit of work. | mini / nano / full |
| Refactoring | Behavior-preserving structural change in small steps. | mini / nano / full |
| Refactoring.Guru | Code smells, refactoring techniques, practical cleanup passes. | mini / nano / full |
| Release It! | Production resilience, timeouts, retries, bulkheads, observability. | mini / nano / full |
| The Pragmatic Programmer | General engineering judgement, automation, feedback loops. | mini / nano / full |
| Working Effectively with Legacy Code | Characterization tests, dependency breaking, safe legacy changes. | mini / nano / full |
Choose by task
Most projects should load one primary rule set. A focused rule beats a pile of overlapping advice.
Code quality
Architecture
Clean Architecture, DDD, DDD Distilled, IDDD, or PoEAA.
Refactoring
Refactoring, APoSD, or Refactoring.Guru.
Legacy code
Production systems
Release It! for failure modes and operations, with DDIA when data semantics matter.
Data systems
FAQ
What are AI coding agent rules and skills?
Instructions that guide coding agents while they generate, review, migrate, or refactor code. Rules are usually always-on or scoped project guidance; skills are task-specific workflows loaded when needed.
Should these be skills or always-on rules?
Use mini and full as skills when the guidance is task-specific. Use nano or one carefully chosen mini file as always-on project context.
Can I use these with Cursor or Claude Code?
Yes. The files are plain Markdown and can be copied into skills, Cursor rules, CLAUDE.md, AGENTS.md, or custom instructions.