openfda-database
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves records from the external openFDA API (api.fda.gov). Although this is a trusted government source, the ingestion of external data that is subsequently processed by the agent creates a potential surface for indirect prompt injection if the records contain adversarial instructions.\n
- Ingestion points: Data is fetched from https://api.fda.gov via the scripts/openfda_query.py script and stored in JSON format.\n
- Boundary markers: The skill does not explicitly instruct the agent to use delimiters or 'ignore' directives when analyzing the contents of the fetched FDA records.\n
- Capability inventory: The skill allows network requests to the FDA domain and file-writing capabilities to the local filesystem.\n
- Sanitization: Raw API responses, including descriptive narrative fields, are written directly to local files without content-level sanitization or filtering.
Audit Metadata