building-apis
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of RESTful endpoints that ingest untrusted data from external callers, presenting a surface for indirect prompt injection. \n
- Ingestion points: Incoming HTTP request bodies, query strings, path parameters, and headers are ingested as source records as described in
SKILL.mdandapi-request.yml. \n - Boundary markers: While
bodySchemais referenced for structural validation inapi-request.yml, there are no instructions for implementing delimiters or warnings to prevent the agent from executing instructions embedded within the data. \n - Capability inventory: The skill documents the ability to execute custom JavaScript handlers (
script.yml), pre/post-map hooks (api-response.yml), and perform read/write operations across various integrations using exports and imports (router.yml). \n - Sanitization: Mappings and input filters provide field extraction and logic gating but do not specifically address the sanitization of natural language instructions within the payload.
Audit Metadata