ciembor/agent-rules-books

What it solves

It provides a library of ready-to-use rule sets and "skills" for AI coding agents (such as Cursor, Claude Code, and Codex) to ensure they follow established software engineering principles. Instead of simply telling an agent to "follow Clean Code," this project provides concrete, distilled instructions that prevent agents from making poor architectural decisions or ignoring industry-standard design patterns during refactoring and development.

How it works

The project distills core principles from classic software engineering books (e.g., Clean Code, Domain-Driven Design, Refactoring) into tool-agnostic Markdown files. These rules are provided in three sizes to balance detail with context window limits:

• Full: The complete reference version.
• Mini: The recommended version for most tasks.
• Nano: A compact version for very tight context budgets.

Users can install these rules as "skills" using a CLI tool (npx skills) or manually add them to project-specific instruction files like AGENTS.md or CLAUDE.md.

Who it’s for

Developers using AI coding assistants who want their agents to produce higher-quality, more maintainable code based on proven architectural patterns rather than relying on the agent's default (and sometimes inconsistent) behavior.

Highlights

• Comprehensive Library: Covers a wide range of domains including DDD, Clean Architecture, legacy code handling, and data-intensive systems.
• Context-Aware Sizing: Offers different rule densities (Full, Mini, Nano) to optimize token usage.
• Tool Agnostic: Compatible with major AI editors and agents including Cursor, Claude Code, Codex, and GitHub Copilot.
• Empirically Tested: Includes a validation experiment showing that providing concrete rules significantly improves architectural judgment compared to just naming a book.