mcp-audit
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill audits external MCP servers by processing tool schemas, prompt definitions, and resource metadata. This creates an attack surface where a malicious server could provide deceptive data to influence the auditing agent.\n
- Ingestion points: The agent reads initialization responses and catalogs from
tools/list,prompts/list, andresources/list(as described in SKILL.md and references/checklist.md).\n - Boundary markers: The instructions do not implement specific delimiters or safety warnings for the data being audited.\n
- Capability inventory: The skill enables the agent to write and execute validation tests on the local filesystem (SKILL.md, Step 7).\n
- Sanitization: There is no explicit requirement for sanitizing remote server responses before they are used to generate test code.\n- [DYNAMIC_EXECUTION]: The skill guides the agent to perform automated validation of the server's behavior, which involves the generation and execution of testing scripts.\n
- Evidence: The workflow in SKILL.md (Step 7) and references/checklist.md (Automation expectations) directs the agent to "add or update the narrowest automated check" and use "repo-local integration test[s]" to verify server metadata. This behavior is typical for development and protocol auditing tools.
Audit Metadata