matlab-analyze-spectrum

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted signal data and descriptions provided by the user, which are used to inform analysis techniques and parameters.
  • Ingestion points: User-provided signal variables, files, and natural language descriptions as described in SKILL.md and references/preliminary-analysis.md.
  • Boundary markers: Absent; there are no explicit delimiters or instructions for the agent to ignore potentially malicious embedded content within the processed data or descriptions.
  • Capability inventory: The skill utilizes evaluate_matlab_code and run_matlab_file to execute complex spectral analysis workflows.
  • Sanitization: Absent; while the skill includes technical diagnostics (e.g., NaN/Inf checks, sampling jitter analysis), it does not implement sanitization to prevent the interpretation of data as instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 08:36 AM
Security Audit — agent-trust-hub — matlab-analyze-spectrum