matlab-define-toolbox-api

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is designed for standard MATLAB development workflows and does not contain malicious instructions, obfuscation, or persistence mechanisms.
  • [DATA_EXFILTRATION]: The skill uses built-in MATLAB functions (dir, which, exist) to inventory files within a local directory specified by the user. No sensitive file access or network exfiltration patterns were detected.
  • [PROMPT_INJECTION]: The skill possesses a data ingestion surface that processes user-provided MATLAB files. 1. Ingestion points: Recursive scanning of the user-provided directory (SKILL.md, Step 2). 2. Boundary markers: The agent relies on keyword detection rather than explicit delimiters. 3. Capability inventory: Uses directory listing, dependency tracing, and local file writing. 4. Sanitization: A mandatory user confirmation step is required before the agent generates the final artifact, providing human oversight.
  • [COMMAND_EXECUTION]: No unauthorized shell commands or remote code execution patterns were found. The skill generates a MATLAB script as an output.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 08:33 PM
Security Audit — agent-trust-hub — matlab-define-toolbox-api