ai-engineering-workflow
AGENTS.md0 starsChanged 4 months agoMITView on GitHub
Read by AI coding assistants (Claude Code, Cursor, Copilot, etc.). This is the single source of truth for how to work in this project. Tool-specific files (.cursorrules, CLAUDE.md) should redirect here rather than duplicate rules. Load the right doc for the task instead of reading everything: - DRY, KISS, SOLID — with judgment, not dogma. Avoid premature abstraction; a wrong abstraction costs more than a little duplication. Apply SOLID's spirit (single responsibility, composition) over ceremony. - Tests are the guardrail.…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,700 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository6 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- commitSkill
- improve-skillSkill
- pr-descriptionSkill
- safe-rolloutSkill
- scaffold-moduleSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
DRY, KISS, SOLID — with judgment, not dogma: Avoid premature abstraction.
- Very specific: line 44
Never commit to main. Branch names follow the convention in docs/GIT_HOOKS.md.
- Distinctive: line 46
Run lint + typecheck before staging.
- Very specific: line 47
Never commit automatically — only on explicit user authorization.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI Agent Rules
Read by AI coding assistants (Claude Code, Cursor, Copilot, etc.). This is the single source of truth for how to work in this project. Tool-specific files (
.cursorrules,CLAUDE.md) should redirect here rather than duplicate rules.
Documentation Index
Load the right doc for the task instead of reading everything:
| Topic | File | When to read |
| System architecture & layout | docs/ARCHITECTURE.md | Understanding where things live; adding a module or agent tool |
| Git hooks & branch naming | docs/GIT_HOOKS.md | Before committing or pushing |
| Development setup & tooling | docs/DEVELOPMENT.md | First-time setup, tooling questions |
| React & component standards | docs/REACT_GUIDELINES.md | When creating or structuring components |
Tech Stack
- TypeScript, React, Next.js (App Router), Node.js
- Testing: Vitest + @testing-library/react
- Package manager: pnpm (never npm/yarn)
Principles (always apply)
- DRY, KISS, SOLID — with judgment, not dogma. Avoid premature abstraction; a wrong abstraction costs more than a little duplication. Apply SOLID's spirit (single responsibility, composition) over ceremony.
- Tests are the guardrail. New behaviour ships with tests; AI-generated code must prove itself.
- Derive state during render; don't store computed values you can compute.
- No barrel imports — use direct paths to keep bundles lean.
Workflow (spec-driven)
- Brainstorm → 2. Spec (specs/) → 3. Implementation plan (specs/implementation-plans/) → 4. Implement.
For agentic work, use the Superpowers sub-skills (superpowers:executing-plans,
superpowers:subagent-driven-development) to implement plans task-by-task.
Commit & PR rules
- Never commit to
main. Branch names follow the convention in docs/GIT_HOOKS.md. - Run lint + typecheck before staging.
- Never commit automatically — only on explicit user authorization. See .claude/skills/commit/SKILL.md.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.