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/deliciousbuding/codex-browser-bridge/agents-mdgit clone --depth 1 https://github.com/DeliciousBuding/codex-browser-bridgeWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/deliciousbuding/codex-browser-bridge/agents-md)<a href="https://agentmods.dev/instructions/deliciousbuding/codex-browser-bridge/agents-md"><img src="https://agentmods.dev/badge/instructions/deliciousbuding/codex-browser-bridge/agents-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.01730 | $0.01730 |
| Opus 5 | $0.00865 | $0.00865 |
| Sonnet 5 | $0.00346 | $0.00346 |
| Haiku 4.5 | $0.00173 | $0.00173 |
Grade A, and why
codex-browser-bridge 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 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.
How it starts
The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
禁止提交 .env、凭据、私钥。提交前 git diff --staged 自查。
仓库级 Skill
本仓库包含项目级 skill:skills/codex-browser/SKILL.md
该 skill 是 LLM agent 使用全部 52 个 MCP 工具的操作手册,包含工具分组速查、常用工作流、工具选择原则。支持项目级 skill 的 MCP 客户端会自动加载。
MCP 工具设计规范
新增 MCP 工具遵循以下模式:
1. Browser 层(src/browser.rs)
- 每个 CDP 操作封装为一个
pub async fn,接受&Client+ 参数 - CDP 响应解析用私有函数,返回
Result<T> - 使用
#[derive(Deserialize)]私有结构体解析 CDP 响应 execute_cdp_generic()是显式低风险方法 allowlist 的 raw CDP 入口;不要用域名前缀放宽,高风险能力必须走专用受控工具
2. MCP 层(src/mcp/ 目录)
types.rs:ToolHandler枚举新增 variant(当前 52 个工具)handlers.rs:handle_tool_call新增 match arm +handle_*方法schema.rs:registered_tools()新增工具定义profiles.rs:如需加入 profile,更新BASIC_TOOLS/NETWORK_TOOLS数组
3. 安全层(src/security.rs)
- 新增文件操作需经过
validate_file_path()路径穿越防护 - 新增 URL 参数需经过
validate_url()scheme 检查
4. 诊断(src/doctor.rs)
run_diagnostics()— 独立于 MCP Server 的 pipe 探活逻辑
5. 测试
- Extractor 测试(
src/mcp/types.rs):required_str、required_string_vec等 - Schema 测试(
src/mcp/schema.rs):验证工具 required 字段、name order、type=object - 集成测试(
tests/browser_api.rs):公开 browser/helper 行为与安全边界 - Reconnect/mock 测试(
src/client.rs,cfg(not(windows))):tokio::io::duplex()mock pipe - Harness 计划:新的真实/模拟 E2E 应放在
tests/,不要使用#[path]嵌入源码文件 - Mock pipe / E2E harness 等待必须 bounded:正向 pipe read 和 spawned task join 用明确 timeout,负向“不得发 pipe 请求”断言用短 grace window;live E2E 的 doctor preflight 和 cleanup MCP 调用也必须有 timeout。
6. 工具数量
- 当前:52 个 MCP 工具
- 新增工具命名:
codex_<domain>_<action>,group tag 放描述开头
构建
cargo check --locked # 快速检查
cargo test --locked # 全量测试
cargo clippy --locked -- -D warnings # lint
cargo build --locked --release # 发布构建 → target/release/codex-browser-bridge.exe
npm --prefix npm test # npm installer/package helper 测试
.\scripts\live-e2e.ps1 # 可选真实 Codex Desktop + Chrome 冒烟测试
Release 时遵循 docs/release-process.md:确保 Cargo.toml 和 npm/package.json 版本号与 tag 一致,CHANGELOG.md 有对应段,npm 包 dry-run 包含 skills/codex-browser/SKILL.md 和 examples/ 多客户端模板。npm 发布走 Trusted Publishing/OIDC,不在 GitHub secrets 中放 npm write token。
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.
- 3d ago First seen · 109 lines · 1,730 tokens per session scan A 8c2bd9c7523d
codex-browser-bridge AGENTS.md is an instructions file published in the GitHub repository DeliciousBuding/codex-browser-bridge (15 stars, last pushed 15d ago), licensed MIT. It adds 1,730 tokens to every session, about $0.0086 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 2026-08-30.
Other instructions, from other repositories
opencode-webmcp AGENTS.md
AGENTS.md instructions for enihcam/opencode-webmcp, covering webmcp-bridge, key facts, chrome requirement, running and architecture.
grok-chrome-bridge AGENTS.md
AGENTS.md instructions for davidsolheim/grok-chrome-bridge, covering grok chrome bridge, verify and do not.
obscura AGENTS.md
Instructions for h4ckf0r0day/obscura, covering agents.md, build, rendering and stealth, no rendering, with rustls or stealth and test.
nanobrowser CLAUDE.md
Instructions for nanobrowser/nanobrowser, covering claude.md, project overview, development commands, workspace tips and architecture.
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.