engineering-consultancy-leads
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill includes a command to register a Model Context Protocol (MCP) server from the author's domain (mcp.manypi.com). This is a standard setup procedure for the ManyPI tool and traces back to the vendor's infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze data from external sources such as practice directories, planning records, and tender awards, which introduces a potential surface for indirect instructions.
- Ingestion points: External websites and public records mentioned in the sources field of the ManyPI brief in SKILL.md.
- Boundary markers: No specific delimiters or instructions are provided to the agent to ignore potentially malicious content embedded in the ingested lead data.
- Capability inventory: The skill configures a data extraction brief for an external tool but does not grant direct shell access or local file system writes within the skill instructions.
- Sanitization: The instructions do not define any specific sanitization or validation steps for the content retrieved from external sources.
Audit Metadata