sobol-decomposition
Pass
Audited by Gen Agent Trust Hub on Jun 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, data exfiltration, or unauthorized command execution detected. The skill logic is strictly focused on mathematical variance-based sensitivity analysis using Sobol sequences and Saltelli sampling.
- [PROMPT_INJECTION]: The skill processes a model_description input which serves as an indirect injection surface for the spawned subagent. 1. Ingestion points: model_description in prompt.md. 2. Boundary markers: Absent. 3. Capability inventory: The skill specifies execution via subagent using spawn-agent which grants tool access. 4. Sanitization: No explicit validation or delimiting of the input is performed. This represents a standard capability surface for processing external model definitions and does not indicate malicious activity.
Audit Metadata