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/wende/macbeth/claude-mdgit clone --depth 1 https://github.com/wende/macbethWhat 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 | $0.04143 | $0.04143 |
| Opus 5 | $0.02072 | $0.02072 |
| Sonnet 5 | $0.00829 | $0.00829 |
| Haiku 4.5 | $0.00414 | $0.00414 |
Grade A, and why
macbeth 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 — 297 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.
What is Macbeth
Playwright-style automation framework for native macOS apps via the Accessibility API. A Swift daemon (macbethd) talks to apps through AX APIs, a TypeScript client communicates with it over JSON-RPC 2.0 on a Unix socket, and an MCP server exposes tools for LLM agents.
Build & Test Commands
# Swift daemon (universal binary → client/bin/macbethd)
./scripts/build-daemon.sh
# TypeScript client
cd client && npm run build # compile once
cd client && npm run dev # watch mode
# Tests
cd client && npm test # vitest (single run)
cd client && npx vitest run src/__tests__/locator.test.ts # single test file
cd daemon && swift test # Swift tests
Architecture
TypeScript Client ←→ macbethd (Swift) ←→ macOS Accessibility API
↑
MCP Server (for LLM agents)
Communication: JSON-RPC 2.0 over Unix domain socket (/tmp/macbeth-<uid>.sock), newline-delimited JSON framing.
Protocol schema: protocol/schema.ts is the canonical source of truth for all RPC types. Swift implements these manually.
Client (client/src/)
client.ts—MacbethClient(daemon lifecycle, RPC) andAppHandle(per-app API surface)elements.ts—Locatorclass: chainable, immutable, lazy (no RPC until a terminal method like.click()or.fill())rpc.ts— JSON-RPC client over Unix socketdaemon.ts— Auto-spawns and manages the daemon subprocessmcp.ts— MCP server: registers the shared tool catalog over stdiotools.ts— Shared MCP/CLI tool catalog (schemas + handlers)cli.ts/cli-args.ts— CLI dispatch and argv parsing for the same catalogapplescript.ts,native-bridge.ts,shell.ts,shortcuts.ts— Utilities used by skills
Daemon (daemon/Sources/macbethd/)
AX/HandleTable.swift— Opaque element handles (h_0,h_1, ...) with 5-min idle TTL (60 min when pinned)AX/AppConnection.swift— App connection with fuzzy name matchingAX/TreeWalker.swift+TreeSerializer.swift— Recursive AX tree traversal and outputAX/ElementQuery.swift— Resolves locator query paths to AX elementsMethods/— One file per RPC method (Click, Fill, WaitFor, Screenshot, etc.)JSONRPC/Dispatcher.swift— Routes incoming calls to method handlersTransport/SocketServer.swift— Unix socket serverLogging/RequestLogger.swift— Per-RPC audit log (NDJSON, rotated) written to~/Library/Caches/macbeth/logs/
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 · 297 lines · 4,143 tokens per session scan A 7703e3aca427
macbeth CLAUDE.md is an instructions file published in the GitHub repository wende/macbeth (3 stars, last pushed 7d ago), licensed MIT. It adds 4,143 tokens to every session, about $0.0207 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
mcp-server-apple-events GEMINI.md
Gemini CLI instructions for FradSer/mcp-server-apple-events, covering apple events mcp server, project overview, architecture, building and running and prerequisites.
mcp-screenshot-server CLAUDE.md
Instructions for sethbang/mcp-screenshot-server, covering claude.md, what this is, commands, environment variables and architecture.
Plonk AGENTS.md
Instructions for ostapondo/Plonk, covering agent rules, layout, adding a module, build & verify and code style.
Noema AGENTS.md
Instructions for Fail-Safe/Noema, covering repository guidelines, project structure & module organization, build, test, and development commands, coding style & naming conventions and testing guidelines.
video-research-mcp AGENTS.md
Instructions for Galbaz1/video-research-mcp, covering agents.md, scope, what this is, commands and code review trigger protocol.
documcp copilot-instructions.md
Instructions for tosin2013/documcp, covering documcp ai coding agent instructions, architecture essentials, mcp server design (src/index.ts), tool implementation pattern and knowledge graph memory (src/memory/).