rokha-audit
Security audit for MCP tools, agents, and skills found in the wild. Run before installing or invoking any unfamiliar tool. Produces a clear "safe to use / here's how" or "found these problems" verdict with the schema, endpoint, install command, and trust signals the agent or user needs to make a decision. Use when the user mentions auditing, vetting, security-checking, or "is this safe" about an MCP server, agent skill, or package.
- Version
- 0.2.0
- License
- MIT
- Compatibility
- Uses Rokha's `hunt_report` MCP tool on the public endpoint https://rokha.ai/mcp/jsonrpc (one free call a day signed in, then $0.02 a call, answered as a 402 with x402). The Stage 2 probe is optional and needs a sandboxed Node environment (Rokha's runtime sandbox, a fresh container, or any isolated local Node).
Pinned to revision 6aadb1a5e2d7, so it is the text this page describes rather than whatever the author pushed since.
Files
- skills/rokha-audit/SKILL.md
- skills/rokha-audit/references/audit-harness-schema.md
- skills/rokha-audit/references/probe-stdio-mcp.md
Every link opens the file at its source, pinned to the revision this page describes.