Classic software books distilled for AI agent context windows

What it does This repository translates fourteen classic software engineering books—covering architecture, refactoring, DDD, and legacy code—into drop-in AGENTS.md rules for Codex, Cursor, and Claude Code. Instead of hoping your agent stumbles onto good design, you prepend distilled principles from Fowler, Martin, and others to steer the output.

The interesting bit Every book is released in three context-aware sizes: nano for tight token budgets, mini as the recommended daily driver, and full for canonical reference. The maintainer even catalogs Reddit criticism in CRITICISM.md, which is a level of self-awareness rarely found in prompt-engineering repos.

Key highlights

• Fourteen rule sets drawn from books like Clean Code, Designing Data-Intensive Applications, and Working Effectively with Legacy Code.
• Three tool-agnostic Markdown sizes per book, with exact line, rule, and byte counts generated via standard Unix tools.
• Works across Codex, Cursor, and Claude Code without vendor-specific lock-in.
• Includes a compatibility matrix and a living document of external criticism.

Caveats

• The project is explicitly labeled v0.5, so the rule sets are still settling.
• The README openly points to Reddit criticism rather than claiming infallibility.
• Compression is extreme—Domain-Driven Design drops from 979 lines to 39—so fidelity to the original text is an act of trust.

Verdict A handy resource if you want AI-generated code that respects boundaries, abstractions, and failure modes. If you do not use agent-specific rules or prefer to read the originals, this is just a well-organized shelf of compressed wisdom.