The Agent Skills Directory

[PROMPT_INJECTION]: The instructions include strong directives to prioritize the provided static analysis tool over the model's internal training data. These are architectural guidelines intended to ensure technical accuracy and do not attempt to bypass safety filters or override system-level instructions.
[REMOTE_CODE_EXECUTION]: The skill demonstrates a secure posture by explicitly prohibiting the use of shell commands or MATLAB execution for diagnosis. It mandates the use of a static analysis tool (mcp__matlab__check_matlab_code), which minimizes the risk of executing arbitrary code.
[DATA_EXFILTRATION]: No network operations or data transmission to untrusted external domains were found. All documentation and license references point to official MathWorks domains, which are consistent with the skill's stated authorship.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of MATLAB source code provided by users.
Ingestion points: The contents of .m files are read into the agent's context for analysis.
Boundary markers: No explicit delimiters are used to separate user code from agent instructions.
Capability inventory: The skill is limited to reading code and providing refactoring suggestions. It lacks capabilities for network access, arbitrary file writing, or command execution based on the ingested content.
Sanitization: No sanitization of the input code is performed.
The impact is negligible because the skill is strictly restricted to static analysis and does not execute the code being analyzed.

matlab-diagnose-parfor