coding-agent-kit
AGENTS.md1 starsChanged 39 days agoMITView on GitHub
Single source of truth for every coding agent in this repository. CLAUDE.md and the opt-in pointer files in contrib/untested/ only point here. What must always hold is enforced by hooks, ops/verify.sh and ops/check-agent-kit.sh; this file holds only what an agent cannot infer from the code. Max 7 rules.
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 33,315 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.
- CLAUDE.mdCLAUDE.md
- kit-debugSkill
- kit-exploreSkill
The file
We split this AGENTS.md into the rules it gives an agent
5 rules. About 1 is common, 1 is distinctive, and 3 are very specific to this project.
≈1
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Verify: ./ops/verify.sh — tests, lint and kit check.
- Very specific: line 20
Read docs/agent/STATUS.md first if it is not already in your context, then run ./ops/verify.sh before new work.
- Very specific: line 27
When code changed, rewrite docs/agent/STATUS.md before presenting: what changed, verification: (verified | partial | failed) with the …
- Common: line 31
Never write secrets, credentials, connection strings, personal data or pasted logs into any agent file.
- Distinctive: line 33
Ask before adding dependencies, changing schemas, deleting files, or touching auth, payments or production configuration.
≈ “Common” is an estimate: at least 1 in 5 of the 42,283 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
AGENTS.md
Single source of truth for every coding agent in this repository. CLAUDE.md and the opt-in
pointer files in contrib/untested/ only point here. What must always hold is enforced by hooks,
ops/verify.sh and ops/check-agent-kit.sh; this file holds only what an agent cannot infer from
the code. Max 7 rules.
Project
- Purpose:
<one sentence: what this software does and for whom> - Stack: <languages, frameworks, package manager, runtime versions>
- Layout: <
src/application code,tests/tests,docs/documentation> (match the real tree)
Commands
- Verify:
./ops/verify.sh— tests, lint and kit check; the only definition of "done" - Setup:
<command>· Run locally:<command>
Rules — ranked, max 7
- Nothing is done until
./ops/verify.shpasses and you quote its last lines. Never changeops/verify.sh,ops/check-agent-kit.sh,ops/agent/,.githooks/, hook configs or CI. - Read
docs/agent/STATUS.mdfirst if it is not already in your context, then run./ops/verify.shbefore new work; a failure inherited from the last session comes first. - Classify each change after reading the code. T0 (trivial, under ~20 lines): verify only.
T1 (feature or bug fix): a change file from
docs/agent/changes/_template.mdbefore code; a fix starts from the root cause (.agents/skills/kit-debug/SKILL.md) and needs a test that failed before it. T2 (schema, auth, payments, cross-module): T1 plus Design and Spec delta. Unsure: the higher tier. Done: close it, run./ops/agent/changes.pyarchive. - When code changed, rewrite
docs/agent/STATUS.mdbefore presenting: what changed,verification:(verified | partial | failed) with the command and commit, what is next. Under 60 lines. - Append chosen-between alternatives to
docs/agent/DECISIONS.md; propose repeated lessons indocs/agent/LEARNINGS.mdwith evidence. Never delete or rewrite a past decision. - Never write secrets, credentials, connection strings, personal data or pasted logs into any agent file. Environment variables by name; people by role.
- Ask before adding dependencies, changing schemas, deleting files, or touching auth, payments or production configuration.
Conventions (only what no linter enforces)
- <e.g. "All HTTP handlers return errors through src/errors, never raw exceptions">
Git
- Imperative subject ≤ 72 characters, one change per commit. Never push directly to
main: every change goes through a branch and a pull request. - Records only (every changed path is in
ops/agent/non-code-paths.txt): at wrap-up, once./ops/verify.shpasses, title the pull request "Records: …" and merge it yourself when CI passes. Anything else waits for the maintainer's merge.
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.

