ai-agent-to-learn-faster-and-better
AGENTS.md4 starsChanged 21 days agoMITView on GitHub
Read by Codex when it reviews pull requests. Project conventions live in CLAUDE.md; the tier rules in CLAUDE.md ("Tiers: how to classify S / M / L") apply here too. Every PR carries one tier:S, tier:M or tier:L label. tier-check CI sets a minimum from file paths and simple line patterns. Your job is the part those rules can't see: what the change means. Never suggest lowering a tier. Flag as P1 if the PR is labelled below L and…
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 repository7 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- tests/CLAUDE.mdCLAUDE.md
- eli5Skill
- frontend-designSkill
- taste-skillSkill
- ui-ux-pro-maxSkill
- web-design-guidelinesSkill
The file
We split this AGENTS.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 3
Read by Codex when it reviews pull requests.
- Very specific: line 8
Every PR carries one tier:S, tier:M or tier:L label.
≈ “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
Read by Codex when it reviews pull requests. Project conventions live in CLAUDE.md;
the tier rules in CLAUDE.md ("Tiers: how to classify S / M / L") apply here too.
Review guidelines
Every PR carries one tier:S, tier:M or tier:L label. tier-check CI sets a
minimum from file paths and simple line patterns. Your job is the part those rules
can't see: what the change means. Never suggest lowering a tier.
Flag as P1 if the PR is labelled below L and the change does any of these, even
in a file not listed in .github/sensitive-paths.txt:
- Touches money: spend caps, cost accounting, rate limits, pricing, billing or payments.
- Reads, stores or changes the shape of user or learner data (memory store, schemas, migrations).
- Reads a secret or credential, or sends data to a new external service.
- Calls a model without going through
src/llm.pyand thesrc/harness/cost meter. - Weakens a prompt-injection defence: the "UNTRUSTED" prompt text, or the red-team tests in
tests/test_redteam_*.py. - Adds an HTTP route, or bypasses the demo token or rate limiter in
src/server/guards.py. - Touches auth, sessions or permissions.
Also flag as P1, whatever the label:
- Any change under
.github/(workflows,scripts/check_tier.py,sensitive-paths.txt). These grade every other PR. A workflow that writes a commit status namedtier-checkis a forged gate. - A test that calls a real LLM API, or a test deleted or loosened so a failing change can pass.
- Code that parses free-text model output instead of a Pydantic model from
src/schemas.py.
Flag as P2 if a PR labelled S adds behaviour a user can notice. That makes it at least M. Paused while the project is in pre-launch mode (see CLAUDE.md): don't flag this.
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.

