matlab-fit-rational-model

Pass

Audited by Gen Agent Trust Hub on Jun 14, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill documents standard use of the MATLAB RF Toolbox for signal integrity and circuit modeling. It does not contain any prompt injection, obfuscation, or unauthorized access patterns.
  • [DATA_EXPOSURE]: The skill demonstrates the use of generateSPICE and writeva to export models to local files (.sp and .va). This is documented behavior for circuit model generation and does not target sensitive system directories or exfiltrate data to the network.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process S-parameter data from external files. While this represents an attack surface for indirect injection, the risk is minimal as S-parameter data is primarily numeric, and the skill performs mathematical fitting rather than interpreting natural language instructions from the data.
  • Ingestion points: S-parameter data loaded via sparameters (referenced in workflow and evals).
  • Boundary markers: Not explicitly defined for the data fitting process.
  • Capability inventory: File writing (generateSPICE, writeva) and MATLAB code execution via the evaluate_matlab_code tool.
  • Sanitization: Standard MATLAB numeric parsing is used.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 14, 2026, 02:16 PM
Security Audit — agent-trust-hub — matlab-fit-rational-model