drug-safety-specialist
Pass
Audited by Gen Agent Trust Hub on Mar 31, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection because it is designed to ingest and process untrusted external data.
- Ingestion points: The skill processes adverse event (AE) reports from 'spontaneous reports', 'clinical trials', 'literature', and 'real-world data' as specified in the thinking patterns of SKILL.md.
- Boundary markers: The system instructions do not provide explicit boundary markers or instructions to treat data from these external sources as non-executable text, which could lead to the agent following instructions embedded within reports.
- Capability inventory: The skill performs medical coding, causality assessment, and regulatory report generation based on the ingested data.
- Sanitization: There is no evidence of sanitization, validation, or escaping of the external content before it is incorporated into the agent's processing context.
- [EXTERNAL_DOWNLOADS]: The skill references multiple external documentation sources from well-known regulatory organizations.
- Evidence: Links point to official ICH guidelines on database.ich.org and regulatory resources on fda.gov and ema.europa.eu (found in SKILL.md and references/6-domain-knowledge.md). These are legitimate and trusted sources for the medical and pharmaceutical domain.
Audit Metadata