Breaking

Agent-rules-books turns 13 software design books into AGENTS.md rules

July 6, 2026 at 01:48 EDT

Software Dev & Coding
AI Agents
Open Source

An open-source project called agent-rules-books, which reworks the design principles from 13–14 classic software engineering books such as Clean Code and Domain-Driven Design (DDD) into AGENTS.md-style rule sets for AI coding agents, has been released. Created by Maciej Ciemborowicz (ciembor on GitHub), it is offered under the MIT license and is aimed at popular AI coding tools including Codex, Cursor and Claude Code.

v0.5 · agent-rules-books · MIT License

Classic software engineering, compressed into rules AI agents can read

An open-source project distills timeless design wisdom from Clean Code, DDD, Clean Architecture and more into AGENTS.md rule files — teaching AI coding tools to write maintainable code while conserving precious context.

13–14

classic engineering books distilled into rules

~2,000

GitHub stars and growing

tiers per book: full · mini · nano

The Clean Code trade-off, drawn to scale

Rules per version — same book, compressed for tighter context windows.

220

full

297 lines

mini

47 lines

nano

32 lines

Recommended: mini — enforces consistent design at a fraction of the context cost.

How it works

Pick a book & tier

→

Copy into project root as AGENTS.md

→

Agent consults rules while generating code

Only rules that actually change an agent's decisions are kept — redundant and duplicate rules stripped out, not copied verbatim.

Why developers value it

Enforces consistent design across AI-generated code
Conserves context via compressed tiers
Practical mini versions ready to drop in

The open questions

Too many rules risk context overflow
Impact depends on how the AI interprets them
Caveats collected in CRITICISM.md

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.