Please report security vulnerabilities through GitHub's private vulnerability reporting rather than opening a public issue. We aim to acknowledge a report within 48 hours and will coordinate a fix and disclosure timeline with you.

In scope: the build pipeline (scripts/build-engine.mjs, scripts/sync-corpus.mjs) and the static site (site/) — anything that could let a crafted input or a compromised build step run unintended code, either at build time or in a visitor's browser.

Not a vulnerability: the playground lets a visitor paste an arbitrary MCP JSON-RPC frame to see the guard's verdict on it, including deliberately malicious frames drawn from the same corpus as getmcpm/mcp-guardbench. The whole point is to try attack payloads against the engine; a payload producing an unexpected verdict is a detection question, not a vulnerability in this site. Everything runs client-side in the visitor's own browser — no server executes or stores what is pasted.

This is a static site with no versioned releases; only main is supported.

getmcpm/cli's SECURITY.md documents the disclosure process for the mcpm CLI itself, whose guard engine this playground runs (pinned per engine.lock.json).