The Agent Skills Directory

[SAFE]: The skill is authored by MathWorks and provides a structured interface for MATLAB and Simulink operations. It follows least-privilege principles by requiring user consent before saving files to disk and by avoiding arbitrary command execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied model names and configuration parameters that are passed into the MATLAB environment. While this is an ingestion surface for untrusted data, the instructions require mapping inputs to specific function arguments via a controlled wrapper (configure_for_codegen), which mitigates the risk of direct command injection.
Ingestion points: User-provided model names and configuration preferences in SKILL.md rules.
Boundary markers: Parameters are passed as structured arguments to a MATLAB function rather than through direct string interpolation of shell commands.
Capability inventory: The skill uses evaluate_matlab_code to interact with the MATLAB environment, allowing it to modify model settings and generate source code artifacts.
Sanitization: The skill instructions indicate that the underlying script performs validation, such as stripping file extensions and checking for model availability, before execution.
[DYNAMIC_EXECUTION]: The skill executes a pre-compiled MATLAB P-code file (configure_for_codegen.p) via the evaluate_matlab_code tool. This is the standard, expected method for distributing protected MATLAB logic and does not represent malicious obfuscation.
[DATA_EXPOSURE]: The skill generates configuration reports and code artifacts. Security is maintained by placing these files in the model's local directory by default and requiring explicit user authorization before any in-memory changes are persisted to the filesystem.

simulink-configure-model-for-code-generation