scanpy
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [SAFE]: All provided code snippets and instructions are consistent with the primary purpose of analyzing biological data. No suspicious network requests, file system modifications, or credential handling were detected.
- [METADATA_POISONING]: The skill metadata contains a 'risk: critical' tag which is misleading given the skill's actual safe content, likely due to a template error.
- [COMMAND_EXECUTION]: The skill references a bundled script (scripts/qc_analysis.py) for automated analysis, which is a standard pattern for agent tools.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data via sc.read_h5ad and other functions. 1. Ingestion points: File reading commands in SKILL.md. 2. Boundary markers: Not specified. 3. Capability inventory: Use of scanpy tools and shell-based script execution. 4. Sanitization: Not specified.
Audit Metadata