agentleFS
kapadiaskapadias/nonna/hosts/AGENTS.md

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…

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
  1. Distinctive:

    Context is a budget: Keep the always-on surface tiny.

    line 17
  2. Very specific:

    Do not skip stages. One exception: a trivial, reversible fix may take the fast lane (.claude/skills/fix/SKILL.md) — check-trivial.sh …

    line 27
  3. Distinctive:

    Only then: the minimum code that works.

    line 40
  4. Very specific:

    Commit or push to main/develop, or force-push.

    line 46
  5. Common:

    Put a secret in code, logs, traces, or prompts.

    line 47

≈ “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.

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

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

  1. 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
  2. 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
  3. 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 → Sync

Do 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:

  1. YAGNI — does it need to exist at all?
  2. Already in this codebase? Reuse it.
  3. The stdlib does it? Use it.
  4. A native platform feature covers it? Use it.
  5. An already-installed dependency solves it? Use it.
  6. Can it be one line? One line.
  7. 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

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.