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.mdandreferences/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_codeandrun_matlab_fileto 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