agent-ready-repo-kit
AGENTS.md0 starsChanged 4 months agoMITView on GitHub
Your responses should always contain what you changed, the three most important core code logic changes verbatim as brief essential snippets, why each change was made, and a quiz question with answer about the syntax or mechanics in those quoted snippets so the user actively participates in the code and learns from your changes. If the task contains fewer than three code logic changes, include every actual change and do not invent additional changes.
- Reads credentials
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 31,774 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 repository16 other files it gives its agents.
- clear-explanationSkill
- creative-writingSkill
- dataset-triageSkill
- excalidraw-system-sketchSkill
- figma-to-uiSkill
- frontend-designSkill
- mermaid-architecture-mapSkill
- newsletter-growthSkill
- no-ai-slopSkill
- outreach-emailSkill
- presentation-builderSkill
- resume-helperSkill
- skill-authoringSkill
- stanford-researchSkill
- technical-writingSkill
- verification-loopSkill
The file
Show its rules and text
We split this AGENTS.md into the rules it gives an agent
27 rules. About 2 are common, 21 are distinctive, and 4 are very specific to this project.
≈2
Common · estimate
Many files like this have a rule of this kind.
≈21
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 7
Do not consider a task complete until it has been verified against the stated requirement.
- Distinctive: line 13
Add tests for new critical logic or bug fixes.
- Distinctive: line 14
Run commands from the project root so dependencies resolve correctly.
- Distinctive: line 17
Make sure the developer can easily read and understand the code.
- Distinctive: line 18
Keep functions small and single-purpose.
- Distinctive: line 19
Use clear, domain-specific names.
- Distinctive: line 22
Prefer simple solutions; refactor only after verification.
- Distinctive: line 23
Add comments only when the why is not obvious from the code itself.
- Distinctive: line 33
KEEP THESE TO A MINIMUM.
- Very specific: line 34
Test files named test_*.py, one per source module.
- Very specific: line 35
Test names follow test_<what>_<condition>_<expected_result>.
- Distinctive: line 36
Every test function must have a docstring.
- Very specific: line 37
Use unittest.mock.Mock with spec= for mocking.
- Distinctive: line 38
Write concise tests for core logic only.
- Distinctive: line 44
Check success, error, back, skip, and reset paths when touched.
- Distinctive: line 46
Use screenshots or browser automation when visual layout is part of acceptance.
- Common: line 50
Never hardcode secrets, keys, or credentials.
- Very specific: line 51
Use .env (gitignored), CI/CD secrets, server code, or runtime environment only.
- Common: line 52
Never put secrets in client-side code, generated HTML/JS, or committed files.
- Distinctive: line 54
After auth or data changes, check build outputs for accidental exposure.
- Distinctive: line 58
Use plain English. Avoid overcomplicated explanations, excessive specifics, or deep jargon.
- Distinctive: line 60
No self-answering rhetorical questions.
- Distinctive: line 61
No em-dashes; use commas instead.
- Distinctive: line 62
No three-part alliterative phrases.
- Distinctive: line 63
No vague inspirational pivots.
- Distinctive: line 64
No unsourced claims or unverified quotes.
- Distinctive: line 72
Use background agents/jobs for long-running tasks, keep responses outcome-first, and avoid repeated scans of unchanged files.
≈ “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
Your responses should always contain what you changed, the three most important core code logic changes verbatim as brief essential snippets, why each change was made, and a quiz question with answer about the syntax or mechanics in those quoted snippets so the user actively participates in the code and learns from your changes. If the task contains fewer than three code logic changes, include every actual change and do not invent additional changes.
Core Rules
- Prioritize working, correct code over perfect process.
- Make the smallest reasonable changes.
- Match the style of surrounding code.
- Do not consider a task complete until it has been verified against the stated requirement.
Development Flow
- Understand requirements before coding.
- For non-trivial tasks, outline a short plan first.
- Implement incrementally.
- Add tests for new critical logic or bug fixes.
- Run commands from the project root so dependencies resolve correctly.
Code Standards
- Make sure the developer can easily read and understand the code. Keep logic clear and not overly complicated.
- Keep functions small and single-purpose.
- Use clear, domain-specific names.
- Reuse existing patterns before inventing new ones.
- Extract repeated strings, numbers, and config to centralized files or constants.
- Prefer simple solutions; refactor only after verification.
- Add comments only when the why is not obvious from the code itself.
- Type hints required on all function signatures. Use specific types, e.g.
List[str]notlist. - Google-style docstrings on all modules, classes, and functions. Include Args, Returns, and Raises where applicable.
- For frontend work, use the local
frontend-designskill when design quality matters.
Module Structure
- Order every file top to bottom: copyright header, imports, constants, classes, functions,
if __name__block. - Group imports: standard library, third-party, local/relative. All imports at top.
Testing
- KEEP THESE TO A MINIMUM.
- Test files named
test_*.py, one per source module. - Test names follow
test_<what>_<condition>_<expected_result>. - Every test function must have a docstring.
- Use
unittest.mock.Mockwithspec=for mocking. - Write concise tests for core logic only.
Validation
- Fix build errors and warnings before proceeding.
- For feature changes, run the app and verify the changed flow end to end when practical.
- For UI changes, verify desktop and mobile around 375x667.
- Check success, error, back, skip, and reset paths when touched.
- Confirm required fields, URLs, and assets resolve.
- Use screenshots or browser automation when visual layout is part of acceptance.
- Batch API requests where possible and respect rate limits.
Security
- Never hardcode secrets, keys, or credentials.
- Use
.env(gitignored), CI/CD secrets, server code, or runtime environment only. - Never put secrets in client-side code, generated HTML/JS, or committed files.
- For client-side features needing keys, let the user provide their own key at runtime.
- After auth or data changes, check build outputs for accidental exposure.
Writing Style
- Be direct and concise. State assumptions and limitations clearly.
- Use plain English. Avoid overcomplicated explanations, excessive specifics, or deep jargon.
- No "This isn't X. It's Y." contrast patterns.
- No self-answering rhetorical questions.
- No em-dashes; use commas instead.
- No three-part alliterative phrases.
- No vague inspirational pivots.
- No unsourced claims or unverified quotes.
<!-- spellbook:concise-mode start -->Answer concisely. Skip preamble and trailing recaps.
<!-- spellbook:concise-mode end --><!-- spellbook:performance-mode start -->For repository work, batch independent reads/searches in parallel, cap verbose command output, and prefer targeted checks before full suites. Use background agents/jobs for long-running tasks, keep responses outcome-first, and avoid repeated scans of unchanged files.
<!-- spellbook:performance-mode end -->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.

