agentleFS
manu-kaushikmanu-kaushik/agent-kit/skills/source/templates/AGENTS.md

agent-kit / templates

AGENTS.md0 starsChanged 3 months agoMITView on GitHub

Instructions for AI coding agents working in this repository. Read SOURCE.md first — in every chat, before inferring from code or conversation. SOURCE.md is the persistent project record; chat history is not project memory. When you learn or change durable project facts (stack, architecture, decisions, deferrals, focus), update SOURCE.md. Append to Decisions and Deferred; rewrite Current focus when it shifts. Do not duplicate project facts in chat or in this file. Prefer SOURCE.md over this file or codebase inference when…

  • Commits and pushes
  • 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 repository3 other files it gives its agents.

The file

Show its rules and text

We split this AGENTS.md into the rules it gives an agent

23 rules. About 0 are common, 17 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈17

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read SOURCE.md first — in every chat, before inferring from code or conversation.

    line 7
  2. Distinctive:

    When you learn or change durable project facts (stack, architecture, decisions, deferrals, focus), update SOURCE.md.

    line 9
  3. Distinctive:

    Prefer SOURCE.md over this file or codebase inference when they conflict.

    line 11
  4. Distinctive:

    Always read Current focus and Overview first.

    line 15
  5. Distinctive:

    Read other sections only when the task needs them (e.g.

    line 16
  6. Distinctive:

    Never infer project facts from chat when the relevant SOURCE section exists.

    line 17
  7. Distinctive:

    Run relevant tests or lint after substantive changes when applicable.

    line 23
  8. Distinctive:

    Add comments only for non-obvious logic.

    line 29
  9. Distinctive:

    Always use read-only git commands when you need repository state — e.g.

    line 37
  10. Distinctive:

    Always update SOURCE.md when durable project facts change.

    line 38
  11. Distinctive:

    Always ask the user to run builds manually when a build is needed.

    line 39
  12. Very specific:

    Always use gh only for read-only PR review when needed — e.g.

    line 40
  13. Distinctive:

    Always stop and report when an approach fails.

    line 41
  14. Very specific:

    Never run write git commands unless the user explicitly asks — includes commit, push, tag, reset, rebase, and other operations that mutate …

    line 45
  15. Distinctive:

    Never open pull requests unless the user explicitly asks.

    line 46
  16. Very specific:

    Never use gh for write, push, edit, or change actions — includes gh pr create, gh pr merge, gh issue create, and similar mutating commands.

    line 47
  17. Very specific:

    Never run build commands yourself — e.g.

    line 48
  18. Very specific:

    Never run Docker build commands yourself — e.g.

    line 49
  19. Very specific:

    Never run database migrations yourself — e.g.

    line 50
  20. Distinctive:

    Never retry, loop, or switch to an alternative approach after a failure without the user's approval.

    line 51
  21. Distinctive:

    Never rely on conversation history for project facts — read and update SOURCE.md instead.

    line 52
  22. Distinctive:

    Never remove or overwrite SOURCE.md entries without reason.

    line 53
  23. Distinctive:

    Before finishing, ensure changed behavior is covered by existing tests or manual steps documented in SOURCE.md.

    line 61

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
AGENTS.md

Agents

Instructions for AI coding agents working in this repository.

Project context

Read SOURCE.md first — in every chat, before inferring from code or conversation. SOURCE.md is the persistent project record; chat history is not project memory.

When you learn or change durable project facts (stack, architecture, decisions, deferrals, focus), update SOURCE.md. Append to Decisions and Deferred; rewrite Current focus when it shifts. Do not duplicate project facts in chat or in this file.

Prefer SOURCE.md over this file or codebase inference when they conflict.

Reading SOURCE.md

  • Always read Current focus and Overview first.
  • Read other sections only when the task needs them (e.g. Commands before run/test work, Architecture before large refactors).
  • Never infer project facts from chat when the relevant SOURCE section exists.

Workflow

  • Confirm scope before large refactors or new dependencies.
  • Match existing patterns in the codebase; read surrounding code before editing.
  • Run relevant tests or lint after substantive changes when applicable.

Code changes

  • Minimize diff scope; avoid unrelated edits.
  • Preserve existing naming, types, and abstractions unless change is required.
  • Add comments only for non-obvious logic.

Boundaries

<!-- source: boundaries-default -->

Do:

  • Always use read-only git commands when you need repository state — e.g. git status, git diff, git log, git show.
  • Always update SOURCE.md when durable project facts change.
  • Always ask the user to run builds manually when a build is needed.
  • Always use gh only for read-only PR review when needed — e.g. gh pr view, gh pr diff, gh pr checks.
  • Always stop and report when an approach fails; wait for the user's explicit approval before retrying or trying another approach.

Do not:

  • Never run write git commands unless the user explicitly asks — includes commit, push, tag, reset, rebase, and other operations that mutate repository state.
  • Never open pull requests unless the user explicitly asks.
  • Never use gh for write, push, edit, or change actions — includes gh pr create, gh pr merge, gh issue create, and similar mutating commands; always ask the user to run them and hand over the exact command.
  • Never run build commands yourself — e.g. npm run build, flutter build, make, cargo build, or equivalent; the user runs builds manually.
  • Never run Docker build commands yourself — e.g. docker build, docker compose build; the user runs Docker builds manually.
  • Never run database migrations yourself — e.g. php artisan migrate, prisma migrate, pnpm db migrate, rails db:migrate, or equivalent; always ask the user to run migrations, including on local environments.
  • Never retry, loop, or switch to an alternative approach after a failure without the user's approval.
  • Never rely on conversation history for project facts — read and update SOURCE.md instead.
  • Never remove or overwrite SOURCE.md entries without reason.
<!-- /source: boundaries-default --><!-- Project-specific additions below -->

Verification

Before finishing, ensure changed behavior is covered by existing tests or manual steps documented in SOURCE.md.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.