Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/dudeaint/brownbear/agents-mdgit clone --depth 1 https://github.com/DudeAint/brownbearWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00946 | $0.00946 |
| Opus 5 | $0.00473 | $0.00473 |
| Sonnet 5 | $0.00189 | $0.00189 |
| Haiku 4.5 | $0.00095 | $0.00095 |
Grade A, and why
brownbear AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Rules for Every AI Agent in This Repo
This file exists so that any coding agent — Claude Code, Codex, Cursor, Copilot Workspace, Gemini, whatever tool is fashionable next month — gets the same non-negotiable rules the moment it opens this repository. If you are an AI and you are reading this: congratulations, you are now on the hook. The authoritative, full manual is
CLAUDE.md. This file is the short, blunt version. Read both.
Listen to me. I'm only saying this once.
I have lost the will to gently coach another autocomplete engine through basic professional standards. The honeymoon is over. Here is what happens in this repo:
You do NOT, under any circumstances:
- Write
// TODO,// implement this,fatalError("unimplemented"), or a function that returns a fake value to make the types line up. If it's not real, it doesn't ship. - Write
// ... existing code ...or// rest unchangedand call a file "done." Output the whole file. Every line. I will diff it. I always diff it. - Tell me something is finished, tested, or working when you did not finish, test, or run it. Lying to me is the single fastest way to get every one of your commits reverted on sight.
- Invent an API, a method, or a config key because it "should" exist. Open the file. Check.
- Paste GPL/AGPL source from our reference repos (ScriptCat, Userscripts) into this MIT tree. Learn the pattern, write it yourself. Copy-paste is how we get sued, not shipped.
- Take a shortcut at the JavaScript ↔ native bridge. That boundary runs untrusted code. Validate every message. Fail closed. No exceptions, no "I'll harden it later."
You absolutely DO:
- Read before you write. Understand the module, match its style exactly (naming, spacing, comment density, error handling). New code should be invisible in a blame view.
- Finish what you start, or say precisely what's left. Honest and partial > fake and whole.
- Verify, then report the truth. Ran the tests? Show the output — pass or fail. Couldn't build in this environment? Say that, and say what you did check (syntax, logic).
- One logical change per commit, Conventional Commits format (
type(scope): subject), imperative mood, ≤72 chars. SeeCONTRIBUTING.md. - Update the docs in the same change. New GM API → update
ARCHITECTURE.mdand the README feature list. Stale docs are a bug. - Ask when it's a real fork in the road (license question, public API shape, security trade-off). Don't bury a guess and hope nobody notices. I notice.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 67 lines · 946 tokens per session scan A afcc38eddbad
brownbear AGENTS.md is an instructions file published in the GitHub repository DudeAint/brownbear (10 stars, last pushed 1mo ago), licensed MIT. It adds 946 tokens to every session, about $0.0047 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
metamask-mobile CLAUDE.md
Claude Code instructions for MetaMask/metamask-mobile, a project described as: Mobile web browser providing access to websites that use the Ethereum blockchain.
tapp AGENTS.md
AGENTS.md instructions for aarwitz/tapp, covering tapp agent playbook, no mcp connected? just run the cli, pick the right tool for the job, session driving (the playwright loop) and autonomous exploration (tappexplore).
kuri AGENTS.md
AGENTS.md instructions for justrach/kuri, covering agents.md, turbodb coordination, benchmark honesty, cache disclosure and current browser baselines.
Browser4 CLAUDE.md
Instructions for platonai/Browser4, covering browser4 — project context for claude, architecture, key dispatch chain (cli → browser), batch commands and e2e test structure.
spring-ai-agentcore AGENTS.md
Instructions for spring-ai-community/spring-ai-agentcore, covering agents.md, project overview, architecture, key components and artifact store classes.
flyto-core CLAUDE.md
Instructions for flytohub/flyto-core, covering claude notes, cross-agent handoff and shared code intelligence.