mcp-builder
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions involve analyzing and processing external, untrusted sources such as API documentation, GitHub issues, and remote service responses.
- Ingestion points: SKILL.md indicates that the agent will ingest documentation, READMEs, and API responses from external services.
- Boundary markers: The skill includes specific warnings to treat external documents as untrusted inputs and explicitly forbids executing instructions embedded within them.
- Capability inventory: The skill involves designing and testing tools that perform network and file operations, which are exercised during the validation phase.
- Sanitization: Mandatory requirements for data minimization, redaction of sensitive information, and rigorous input/output schema validation are included in the implementation and validation guides.
Audit Metadata