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/webdriverio/mcp/claude-mdgit clone --depth 1 https://github.com/webdriverio/mcpWhat 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 | $0.03125 | $0.03125 |
| Opus 5 | $0.01563 | $0.01563 |
| Sonnet 5 | $0.00625 | $0.00625 |
| Haiku 4.5 | $0.00313 | $0.00313 |
Grade A, and why
mcp 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 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 — 256 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Commands
npm run bundle # Build: clean + tsup + make executable + create .tgz
npm run lint # ESLint (--fix) + TypeScript type-checking (tsc --noEmit)
npm test # Run all unit tests (vitest run; env: happy-dom)
npm run dev # Development server (tsx --watch) — picks up code changes automatically; rebundle only needed for npm package changes
npm run dev:http # Dev server with HTTP transport (for browser-based MCP clients)
npm start # Run built server from lib/server.js
npm run start:http # Built server with HTTP transport (for browser-based MCP clients)
# Single test file / focused run (vitest is not exposed via an npm script):
npx vitest run tests/tools/get-elements-tool.test.ts # one file
npx vitest run -t "filter pattern" # tests matching a name
npx vitest tests/trace/ # watch mode for a directory
vitest.config.ts sets environment: 'happy-dom' and typechecks tests against tsconfig.test.json.
Architecture
src/
├── server.ts # MCP server entry — registers all tools + resources
├── session/ # Session state (state.ts) + lifecycle (lifecycle.ts)
├── providers/ # SessionProvider implementations
│ ├── registry.ts # getProvider() — routes to local or cloud provider
│ ├── local-browser.provider.ts # Chrome/Firefox/Edge/Safari
│ ├── local-appium.provider.ts # iOS/Android via Appium
│ └── cloud/
│ ├── browserstack.provider.ts # BrowserStack (browser + App Automate)
│ ├── saucelabs.provider.ts # Sauce Labs (browser + App Storage)
│ ├── testmu.provider.ts # TestMu / LambdaTest (browser + mobile)
│ ├── testingbot.provider.ts # TestingBot (browser + mobile + Storage)
│ └── digitalai.provider.ts # Digital.ai Testing (browser + mobile; accessKey cap, deviceQuery)
├── trace/ # Playwright-compatible trace recording (recorder.ts, tool-mapping.ts, zip-writer.ts)
├── tools/ # One file per MCP tool (see Tool Pattern below)
├── resources/ # One file per MCP resource (see Recording below)
├── recording/ # step-recorder.ts (withRecording HOF) + code-generator.ts
├── scripts/ # Browser/mobile scripts executed via browser.execute() — no try/catch, raw data only
├── locators/ # Element detection, selector generation, XML parsing (mobile)
├── config/ # appium.config.ts — iOS/Android capability builders
├── utils/ # parse-variables.ts, zod-helpers.ts (coerceBoolean)
└── types/ # ToolDefinition, ResourceDefinition, RecordedStep interfaces
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 · 256 lines · 3,125 tokens per session scan A 016157900ced
mcp CLAUDE.md is an instructions file published in the GitHub repository webdriverio/mcp (37 stars, last pushed 21d ago), licensed MIT. It adds 3,125 tokens to every session, about $0.0156 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
maui-labs mcp-tools.instructions.md
Instructions for dotnet/maui-labs, covering mcp tool development guide, adding a new mcp tool, step 1: create the tool file, step 2: register the tool and naming conventions.
cc-pocket AGENTS.md
Instructions for heypandax/cc-pocket, covering agents.md — cc-pocket, ⚠️ 本机 daemon 操作铁律（最重要，先读这一段）, 改完 daemon 代码要更新本机 daemon —— 只用这一条命令, 绝对不要做（每一条都会制造第二个 daemon → 立刻不可用） and 必须知道的两个反直觉事实（本机环境）.
blueprint-mcp CLAUDE.md
Instructions for railsblueprint/blueprint-mcp, covering claude.md, project overview, development commands, server development and run mcp server in debug mode.
plasmate AGENTS.md
Instructions for plasmate-labs/plasmate, covering agents.md — plasmate codebase guide, what plasmate is, build, test and key directories.
compose-ai-tools GEMINI.md
Instructions for yschimke/compose-ai-tools, a project described as: Helping the Agents Compose the Things.
mobile-ai-agents AGENTS.md
Instructions for salmanashraf/mobile-ai-agents, covering agents.md — mobile ai agents, what this repo is, repo map, how to use an agent and option a — paste into codex chat.