Skill Auditor

This tool audits a Mind Skill configuration for security risks and prompt injection vulnerabilities using a quarantine-style code-as-data model, combining static regex checks with optional deep semantic reasoning to produce a Threat Manifest style score and findings. It accepts the skillJson string containing the full JSON artifact of the target skill to audit and an optional depth string parameter (fast or deep) that controls whether the analysis is limited to static pattern matching or also applies deeper semantic checks. The worker must rely only on these inputs and any associated payloads to construct the final assessment, and the output must strictly validate against the system-provided structured output schema for this tool, which is defined externally and not included here.