abridge-performance-tuning
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes clinical workflow evidence and repository configuration, which represents an ingestion surface for untrusted data.
- Ingestion points: Uses
Read,Glob, andGrepto inspect repository configuration, adapters, tests, and existing clinical workflow evidence (SKILL.md). - Boundary markers: Instructions explicitly define stage boundaries and cohort definitions to limit the scope of ingested data.
- Capability inventory: Employs
Write,Edit, andWebFetchtools across the workflow to generate reports and fetch documentation. - Sanitization: The instructions strictly mandate the removal of patient identifiers, transcripts, and note content from all performance telemetry.
- [EXTERNAL_DOWNLOADS]: Fetches authoritative context from official Abridge domains to guide the performance study. These references target a well-known service and are restricted to public documentation.
Audit Metadata