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/mockzilla/mockzilla-mcp/claude-mdgit clone --depth 1 https://github.com/mockzilla/mockzilla-mcpWrote 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/mockzilla/mockzilla-mcp/claude-md)<a href="https://agentmods.dev/instructions/mockzilla/mockzilla-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/mockzilla/mockzilla-mcp/claude-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 | $0.01567 | $0.01567 |
| Opus 5 | $0.00783 | $0.00783 |
| Sonnet 5 | $0.00313 | $0.00313 |
| Haiku 4.5 | $0.00157 | $0.00157 |
Grade C, and why
mockzilla-mcp CLAUDE.md scanned grade C with 1 finding 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
clean slate: `rm -rf ~/.cache/mockzilla-mcp`. To test the cache How it starts
The opening of the file, as written. The whole thing — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
mockzilla-mcp
Stdio MCP server bridging Claude Desktop / Cursor / etc. to mockzilla. Pure-JS, no build step, ESM, Node ≥ 18.
Layout
bin/cli.js JSON-RPC entry: stdin loop, dispatch, signal handlers.
lib/tools.js The local tool registry (descriptions + handler refs).
lib/install.js check_cli, install_cli, resolveMockzilla, cache helpers.
lib/local.js serve_locally, stop_locally, peek_openapi, child tracking.
lib/discover.js discover_specs (filesystem scan + spec summaries).
lib/docs.js mockzilla_docs_{topics,read,search}; raw GitHub source +
MOCKZILLA_DOCS_DIR override for contributors editing docs.
lib/version.js Bridge version + npm registry update check (bridge_status).
lib/proxy.js Hosted-plane forwarding (uses MOCKZILLA_TOKEN).
lib/util.js Tiny shared helpers (shellEscape).
bin/cli.js is the only file the npm bin entry runs. It owns the
JSON-RPC loop and delegates everything else.
Two planes
- Local plane. Tools that touch the user's machine. Defined in
lib/tools.js, handlers live inlib/install.jsorlib/local.js. Always available — no auth, no token. - Hosted plane. Tools the Django server defines at
app/mcp/tools.py. Proxied through whenMOCKZILLA_TOKENis set. When the bridge sees atools/callfor a name it doesn't recognise locally, it forwards to the hosted endpoint.
Tool names must not collide — local tools always win the dispatch, so a local tool with the same name as a hosted one would shadow it.
Adding a new local tool
There's a registry pattern — lib/tools.js is the single source of
truth for what the agent sees. The actual handler lives in
lib/local.js (or, if it's an install/cache concern, lib/install.js).
1. Write the handler
Add an async function to lib/local.js:
export async function describe_thing(args) {
// Validate args. Throw if invalid; the bridge translates thrown
// errors into MCP `isError: true` results so the agent surfaces
// them to the user.
if (typeof args.target !== "string") {
throw new Error("`target` must be a string");
}
// Do the work. Plain return value becomes the result's JSON text
// content. Keep it small and structured — the agent reads it.
return { target: args.target, observed: someState() };
}
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 · 170 lines · 1,567 tokens per session scan C 5e8f0aa5c2af
mockzilla-mcp CLAUDE.md is an instructions file published in the GitHub repository mockzilla/mockzilla-mcp (5 stars, last pushed 3mo ago), licensed MIT. It adds 1,567 tokens to every session, about $0.0078 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). 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-mux AGENTS.md
AGENTS.md instructions for mcpmux/mcp-mux, covering agents.md, project overview, workspace layout, build & dev commands and testing.
mcp-mux CLAUDE.md
Claude Code instructions for mcpmux/mcp-mux, covering claude.md, codebase search, what is mcpmux, repository structure and mcp-mux (desktop app) - main project.
tubepull-mcp AGENTS.md
AGENTS.md instructions for miketwalker/tubepull-mcp, covering agents.md — how an ai agent should call tubepull, identity, supported platforms, when to use tubepull and handshake.
testmo-mcp CLAUDE.md
Claude Code instructions for strelec00/testmo-mcp, covering claude.md, commands, install dependencies, run the mcp server directly and run via mcp cli (for testing tools interactively).
MCP_Pokemon GEMINI.md
Gemini CLI instructions for Asthanaji05/MCP_Pokemon, covering gemini's scratchpad, project context, current task and agent directives.
cli-continues parsers.instructions.md
Instructions for yigitkonur/cli-continues, covering parser review guidelines, crash safety (critical), required exports, jsonl streaming and tool summarizer.