mcp-integration
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The tools defined in the skill ingest external data, which is an attack surface for indirect instructions. \n
- Ingestion points: Data retrieved from LinkedIn profiles, company profiles, and spreadsheets through tools like
enrich_contactandread_sheet_rows(SKILL.md). \n - Boundary markers: No specific instruction delimiters or 'ignore' commands are documented for processing external data outputs. \n
- Capability inventory: The skill facilitates data retrieval and enrichment via external APIs; no local shell execution or unauthorized file system access is provided. \n
- Sanitization: There is no mention of input validation or output encoding for data returned from the MCP tools. \n- [EXTERNAL_DOWNLOADS]: The configuration guide references official LeadMagic domains (
mcp.leadmagic.io,leadmagic.io) and their public GitHub repository for documentation and setup. These are verified vendor resources.
Audit Metadata