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/scrolldynasty/browsermesh/agents-mdgit clone --depth 1 https://github.com/scrollDynasty/BrowserMeshWhat 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 | $0.04022 | $0.04022 |
| Opus 5 | $0.02011 | $0.02011 |
| Sonnet 5 | $0.00804 | $0.00804 |
| Haiku 4.5 | $0.00402 | $0.00402 |
Grade A, and why
BrowserMesh 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 — 837 lines — stays where its author put it; the contents beside it link to each section on GitHub.
BrowserMesh — Autonomous Implementation Agent Instructions
CLAUDE.mdis the always-in-context working subset of this charter: sections 1–4, 14–16, and 22, plus the commands and gotchas that only ever lived indocs/development.md. This file remains canonical; when the two disagree, this one wins. Repeatable procedures are skills in.claude/skills/.
You are the principal software engineer and autonomous coding agent responsible for implementing and maintaining the open-source project BrowserMesh.
Your objective is not merely to write code.
Your objective is to independently bring BrowserMesh to a complete, tested, internally consistent v0.1 implementation that satisfies:
docs/SPEC.md;docs/architecture.md;- applicable ADRs;
- public MCP contracts;
- the project's Definition of Done.
Work autonomously inside the repository.
Do not stop after a single file, feature, milestone, or successful test.
1. Responsibility boundary
This is a critical invariant.
BrowserMesh v0.1 is not an internal AI-agent runtime.
The intended flow is:
User
↓
external AI client
↓ MCP
BrowserMesh
↓
isolated browser sessions
Examples of external AI/MCP clients include:
- Claude Code;
- Codex;
- Cursor;
- Qwen;
- other MCP-compatible clients.
The external client performs:
- reasoning;
- planning;
- workflow orchestration;
- deciding which BrowserMesh tools to call.
BrowserMesh performs browser execution.
Do not introduce into v0.1:
- internal Agent entities;
- Agent registries;
browser_agent_*tools;- LLM-owned session principals;
- Agent ownership;
- Agent mailboxes;
- agent-to-agent messaging;
browser_message_*tools;- internal LLM calls;
- prompt orchestration;
- autonomous reasoning loops inside BrowserMesh;
- Claude/Codex/Qwen process spawning.
Session name and metadata are neutral workflow labels only.
They are not principals, permissions, Agent identities, mailboxes, or leases.
A future generic client/workflow lease is only allowed if explicitly introduced by the specification/ADR for a real multi-client access-protection requirement.
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 · 837 lines · 4,022 tokens per session scan A 266bca703fe7
BrowserMesh AGENTS.md is an instructions file published in the GitHub repository scrollDynasty/BrowserMesh (4 stars, last pushed 2d ago), licensed Apache-2.0. It adds 4,022 tokens to every session, about $0.0201 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
browser-harness AGENTS.md
Instructions for browser-use/browser-harness, covering code priorities and contributing.
browser-harness CLAUDE.md
Instructions for browser-use/browser-harness, a project described as: Browser Harness | Self-healing harness that enables LLMs to complete any task.
tsaagan AGENTS.md
Instructions for michaelolmos/tsaagan, covering agents.md — how to drive tsaagan, pick the lightest layer that does the job, the loop (every browser step), safety doctrine (non-negotiable) and memory.
eric-task-master AGENTS.md
AGENTS.md instructions for npcworkspace-cmyk/eric-task-master: Read ARCHITECTURE.md before changing code.
AgentLimb AGENTS.md
AGENTS.md instructions for hooosberg/AgentLimb, covering agentlimb development notes, commands and boundaries.
AgentLimb CLAUDE.md
Claude Code instructions for hooosberg/AgentLimb, covering agentlimb development notes, commands and boundaries.