performance-attribution
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements equity, factor, and multi-period attribution models using standard Python and NumPy. Technical review confirms no access to sensitive system resources, network functions, or dangerous command execution patterns.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided numerical weights and returns. Ingestion points: Class constructors in
scripts/performance_attribution.py. Boundary markers: Absent. Capability inventory: No dangerous operations like subprocess or network access across any scripts. Sanitization: Inputs are strictly converted to floating-point numbers via NumPy. Given the lack of actionable capabilities, the surface is assessed as safe.
Audit Metadata