brownbear CLAUDE.md

A repository instruction file that defines the engineering rules and project context an AI coding agent must follow for the BrownBear iOS browser and userscript engine.

In plain words
What is it for?
Use it to guide implementation, code review, file reading, testing, and decisions about how BrownBear's five modules should be changed.
Why use it?
It gives the agent a project-specific source of truth before code changes, including quality expectations, architecture boundaries, and non-negotiable rules.

Instructions file

Install

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.

agentmods
npx agentmods add instructions/dudeaint/brownbear/claude-md
Clone the repo
git clone --depth 1 https://github.com/DudeAint/brownbear
Per session 2,935 This file is loaded in full into every session.
When invoked 2,935 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What 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.

ModelPer sessionOnce invoked
Fable 5 $0.02935 $0.02935
Opus 5 $0.01468 $0.01468
Sonnet 5 $0.00587 $0.00587
Haiku 4.5 $0.00294 $0.00294

Measured 3d ago against content hash 2e1674802671, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

brownbear CLAUDE.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 3d 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.

CLAUDE.md · 217 lines

How it starts

The opening of the file, as written. The whole thing — 217 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md — Engineering Operating Manual for AI Agents

This file is loaded into context at the start of every Claude Code session in this repository. It is the single source of truth for how code is written here. Treat every rule below as a hard constraint, not a suggestion. If a request conflicts with this file, surface the conflict before proceeding.

BrownBear is a production iOS browser + userscript engine. The bar is App-Store-shippable, enterprise-grade code. We do not ship demos.


⚠️ READ THIS BEFORE YOU TOUCH A SINGLE LINE

Look. I have reviewed enough AI-generated and junior-engineer pull requests to last three lifetimes, and my patience is gone. I have watched // TODO: implement later get merged into main. I have seen a "finished" feature that was three stubbed functions and a return nil wearing a trench coat. I have read a 600-line file that ended in // ... rest unchanged when the rest was very much NOT unchanged. I am tired. My laptop has been to the edge of the wall more than once. We are not doing that here.

So before you start generating, internalize this:

  • If you don't understand the file, READ IT. Open it. Read its neighbors. Do not pattern- match a vibe and hallucinate an API that doesn't exist. Guessing is how we get 2am pages.
  • "It compiles" is not "it works." And "it looks done" is not "it's done." If you didn't run it, say you didn't run it. Do not lie to me with confidence.
  • Half-finished and HONEST beats finished and FAKE. If you ran out of room, stop and tell me exactly what's left. I can work with that. I cannot work with a stub you called complete.
  • This app runs untrusted code from the internet. Every shortcut you take at the JS↔native boundary is a CVE with my name on the commit blame. Slow down. Validate the input. Fail closed.
  • No slop. No filler. No "here's a basic implementation to get you started." I didn't ask for a starting point. I asked for the thing. Build the thing.

Read the full file on GitHub · 217 lines

Changes

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.

  1. 3d ago First seen · 217 lines · 2,935 tokens per session scan A 2e1674802671

Subscribe to this mod's changes

brownbear CLAUDE.md is an instructions file published in the GitHub repository DudeAint/brownbear (10 stars, last pushed 1mo ago), licensed MIT. It adds 2,935 tokens to every session, about $0.0147 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.