nonna / hosts
AGENTS.md0 starsChanged 46 days agoMITView on GitHub
This repository runs Nonna. These are the non-negotiables. Each section links to the full rule in .claude/rules/; read it before you act in that area. Git hooks refuse a commit on main, master or develop, a secret in a commit or a push, and a push with a red test suite or a stale docs/STATUS.md; --no-verify is not yours to use. 1. The LLM proposes; deterministic gates decide. An LLM may read, hypothesize, draft, and explain. Tests, types, linters, schema…
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 32,671 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 repository29 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- adrSkill
- api-designSkill
- auditSkill
- code-reviewSkill
- concurrency-performanceSkill
- coverageSkill
- debuggingSkill
- debugSkill
- fast-laneSkill
- fixSkill
- implementSkill
- intakeSkill
- leanSkill
- migration-safetySkill
- nonnaSkill
- observabilitySkill
- planSkill
- refactoringSkill
- releaseSkill
- reviewSkill
- rollbackSkill
- security-reviewSkill
- shipSkill
- supply-chainSkill
- syncSkill
- tddSkill
- tdd-workflowSkill
- testSkill
The file
We split this AGENTS.md into the rules it gives an agent
5 rules. About 1 is common, 2 are distinctive, and 2 are very specific to this project.
≈1
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 17
Context is a budget: Keep the always-on surface tiny.
- Very specific: line 27
Do not skip stages. One exception: a trivial, reversible fix may take the fast lane (.claude/skills/fix/SKILL.md) — check-trivial.sh …
- Distinctive: line 40
Only then: the minimum code that works.
- Very specific: line 46
Commit or push to main/develop, or force-push.
- Common: line 47
Put a secret in code, logs, traces, or prompts.
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 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
Nonna — house rules for this repository
This repository runs Nonna. These are the non-negotiables. Each section links to the full rule in
.claude/rules/; read it before you act in that area. Git hooks refuse a commit on main, master or
develop, a secret in a commit or a push, and a push with a red test suite or a stale
docs/STATUS.md; --no-verify is not yours to use.
The three principles
- The LLM proposes; deterministic gates decide. An LLM may read, hypothesize, draft, and explain. Tests, types, linters, schema checks, and human review decide what merges or acts. Never act on free-text model output that touches production, money, user data, or an irreversible action without a gate between it and the consequence. → boundaries.md
- Safety is lexicographically prior to speed. Risk-reducing actions (revert, halt, roll back, narrow scope) may be automatic. Every risk-increasing action passes a gate or a human. When the safety layer and the fast layer disagree, safety wins. → safety.md
- Context is a budget. Keep the always-on surface tiny; load depth on demand; delegate fan-out reading to subagents and keep the conclusion, not the dump. Thrift never buys out a test, a review, a validation, or a gate. → token-economy.md
The loop
Research & Reuse → Plan → TDD (RED → GREEN → REFACTOR) → Implement → Review → Verify → Commit & PR → SyncDo not skip stages. One exception: a trivial, reversible fix may take the fast lane (.claude/skills/fix/SKILL.md) —
check-trivial.sh decides eligibility, never prose. → dev-process.md
Before writing code
Understand the problem first — trace the real flow — then stop at the first rung that holds:
- YAGNI — does it need to exist at all?
- Already in this codebase? Reuse it.
- The stdlib does it? Use it.
- A native platform feature covers it? Use it.
- An already-installed dependency solves it? Use it.
- Can it be one line? One line.
- Only then: the minimum code that works.
The ladder sizes the solution, never the loop. → engineering.md
Never
- Commit or push to
main/develop, or force-push. (the git pre-commit hook refuses a commit on them.) - Put a secret in code, logs, traces, or prompts. (the git pre-commit and pre-push hooks block it.)
- Mark work done — in a tracker, in
STATUS.md, or in a PR — with failing tests. - Ship production logic with no test that would have failed before it.
- Override a gate's verdict with your own judgment.
A human approves
First promotion to production · anything that widens blast radius (raising a limit, broadening a permission, deleting at scale) · overriding a safety gate · onboarding an external dependency with access to data or money.
Done means the mirrors agree
Tracker · docs/STATUS.md · docs/ADR · branch + PR · the .claude/ index · memory. Merged code with
a drifted mirror is a silent lie about the state of the system. → sync.md
Branch feature|fix|chore|refactor/<id>-<slug> → PR to develop → PR to main. Conventional
commits, one logical change each. → git-workflow.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.

