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/ryn-mic/ai-browser-tester/agents-mdgit clone --depth 1 https://github.com/Ryn-Mic/ai-browser-testerWrote 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/ryn-mic/ai-browser-tester/agents-md)<a href="https://agentmods.dev/instructions/ryn-mic/ai-browser-tester/agents-md"><img src="https://agentmods.dev/badge/instructions/ryn-mic/ai-browser-tester/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 5.1 | $0.02126 | $0.02126 |
| Opus 5 | $0.01063 | $0.01063 |
| Sonnet 5 | $0.00425 | $0.00425 |
| Haiku 4.5 | $0.00213 | $0.00213 |
Grade A, and why
ai-browser-tester 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 4d 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- ai-browser-tester CLAUDE.md — 91% identical, 16 lines differ
How it starts
The opening of the file, as written. The whole thing — 134 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
本文件为后续 Codex / AI agent 在本项目中工作的项目级说明。
项目概述
本项目目标是构建一个 MCP-first 浏览器控制工具,供 Codex、Claude Code 或任何兼容 MCP client 执行自动化测试:
- 浏览器侧通过 Chrome/Edge 扩展或早期油猴脚本注入页面。
- 页面侧把 DOM、可访问性信息、可交互元素和页面事件转换成 LLM 可消费的结构化快照。
- MCP client 负责规划;本项目通过 MCP 暴露受控、可审计、ref-only 的浏览器能力。
- 控制方式优先采用
snapshot -> ref -> action -> result -> snapshot -> assertion的闭环。
主要计划文档在:
docs/ai-browser-tester-plan.md
当前项目状态
第一轮闭环(implementation-order Phase 1–7)已落地:
- 已初始化 pnpm TypeScript monorepo。
packages/protocol:Zod 协议 schemas + 单元测试(JSON-RPC、握手、tab、snapshot、action、result、event、error、redaction、safety)。packages/page-controller:与浏览器无关的 PageController(snapshot/act/getEvents、ref registry、action executor、event collector、snapshot redaction)。apps/bridge:WebSocket handshake、pairing token 校验、协议版本协商、session manager、extension router、policy 层、audit log(JSONL + fill 值遮蔽)、严格BrowserTools、以及与传输解耦的 MCP tool adapter(createMcpToolServer)。apps/extension:MV3 manifest(host 权限仅 localhost/127.0.0.1,无<all_urls>)、background service worker、pairing token WebSocket client、tab registry、page 请求路由、content script + PageController。dev-only automation hook 仅在AI_BROWSER_TESTER_DEV=1编译。tests/:共享e2e/harness.ts+ 8 个 E2E 规格(login、form、todo/modal、open shadow DOM + same-origin iframe、SPA route change、blocked/stale-tab URL、high-risk 确认、内部滚动容器),真实加载 unpacked MV3 扩展跑通snapshot -> ref -> action -> assert -> events,并校验 audit、密码遮蔽、无 evaluate、无<all_urls>、policy fail-closed。docs/:protocol.md、security.md、architecture.md、testing.md已同步到实现现状。
当前优先项:
- 2026-07-10 已完成 production unpacked Chrome + Codex stdio MCP 真实浏览器测试及阻断问题复测,证据见
docs/mcp-live-test-report-2026-07-10.md;用户随后已授权发布@ryn-mic/ai-browser-testernpm MCP 以及包含独立 extension ZIP/SHA assets 的 GitHub Release,扩展签名和商店上架仍搁置。 - Phase 12 snapshot quality/scale controls 已落地:可配置文本/元素上限、landmark 提取、ARIA widget 状态(checked/expanded/selected)、精确重复文本去重、列表/表格/dialog/内部滚动容器语义摘要、显式 truncation/去重元数据(
page.truncated)、遮蔽先于输出/去重/裁剪,以及大页面/密钥密集回归测试。稳定 ref、事件/错误聚合摘要为有意后延项。 - Phase 13 extension installation/permissions model 已落地:unpacked extension 构建包含 options page,README 提供 Chrome/Edge 安装与 pairing token/bridge URL 配置说明;manifest 永久 host permissions 仍仅 localhost/127.0.0.1,无
<all_urls>;非本地站点通过 options page 逐 origin 请求 optional host permission,并用动态 content script registration 注入 page agent。 - Phase 14 observability/diagnostics 已落地:
browser_diagnostics只读工具、SessionManager.diagnostics()会话/标签/策略健康摘要、BRIDGE_ERROR_CATALOG全错误码目录、诊断 URL 遮蔽与 troubleshooting 文档。 - Phase 15 agent regression/release readiness 已落地:
pnpm test:agentscripted agent 回归、AgentScenarioRecorder指标 JSON(成功率/工具调用数/耗时/snapshot 大小/错误分布)、pnpm release:gate发布门禁、兼容矩阵与 release checklist。 - CI(
.github/workflows/release-gate.yml)在 PR 和 push 上运行pnpm release:gate;pnpm smoke:mcp(scripts/Codex-mcp-smoke.mjs)对真实 stdio MCP CLI 做 transport smoke(initialize/tools.list/tools.call diagnostics),见docs/Codex-mcp-smoke.md。 docs/next-phase-plan.md中的 MCP readiness、手动扩展安装、Codex MCP 注册和真实浏览器测试已经完成;CI 打包已落地,HTTP、内置 real-model regression、扩展签名/商店发布仍后置。
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.
- 4d ago First seen · 134 lines · 2,126 tokens per session scan A 15d2511b92be
ai-browser-tester AGENTS.md is an instructions file published in the GitHub repository Ryn-Mic/ai-browser-tester (0 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 2,126 tokens to every session, about $0.0106 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
antigravity-testing-kit GEMINI.md
Instructions for anhtester/antigravity-testing-kit, covering gemini ai - global automation agent rules, git pull restriction rule, browser rules (mandatory), 🖥️ viewport & mode and 🔄 thứ tự debug bắt buộc (playwright mcp).
mcp-ie-migration-vrt AGENTS.md
AGENTS.md instructions for rayven122/mcp-ie-migration-vrt, covering agents.md, file map, architecture, conventions and adding a tool.
qa-use CLAUDE.md
Instructions for desplega-ai/qa-use, covering claude.md, project map and tech stack.
ClaudeCode_Playwright_Demo CLAUDE.md
Instructions for anhtester/ClaudeCode_Playwright_Demo, covering claude ai - global automation agent rules, git pull restriction rule, browser rules (mandatory), 🖥️ viewport & mode and 🔄 thứ tự debug bắt buộc.
playwright-bdd-cursor-template CLAUDE.md
Instructions for m3au/playwright-bdd-cursor-template, covering running tests, test commands, test organization, writing tests and code architecture.
vibe-testing CLAUDE.md
Claude Code instructions for AishwaryShrivastav/vibe-testing, covering vibe test — code-aware browser testing agent, what this project does, architecture, mcp tools (13) and development.