target-prioritization
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security risks or malicious patterns were identified in the codebase. The implementation follows security best practices, such as proper HTML escaping in reports and avoiding shell execution contexts.
- [COMMAND_EXECUTION]: The orchestrator script (
scripts/orchestrate.py) spawns sub-fetchers usingsubprocess.run()safely by passing arguments as arrays rather than shell strings, avoiding shell command injection risks. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits a standard data ingestion surface from text or CSV tables.
- Ingestion points: Ingests external datasets via
--inputfiles inscripts/orchestrate.pyand fetches public genomic data from authoritative endpoints (UniProt, OpenTargets, HPA, PubMed, ChEMBL). - Boundary markers: Data points are populated into fixed markdown tables inside
scripts/aggregate.pywithout special prompt delimiters. - Capability inventory: Performs safe file-writing operations and read-only network queries to official scientific endpoints.
- Sanitization: Gene symbols are split, trimmed, and strictly parameterized into clean list configurations before invocation.
Audit Metadata