rfd4-proteina-commitment-curves
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes metric data from user-supplied CSV files. Ingestion points: scripts/commitment_curve.py and scripts/metric_distributions.py read external CSV data into dataframes. Boundary markers: None are present in the CSV parsing code. Capability inventory: The skill is limited to reading files, writing results to disk, and generating plots with Matplotlib; it contains no subprocess calls, network access, or dynamic code evaluation (eval/exec). Sanitization: The scripts use pandas numeric coercion which filters out non-numeric injection payloads during data processing.
Audit Metadata