The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses tools like model_edit and model_check to perform structural modifications, connectivity checks, and parameter configuration within Simulink, System Composer, Stateflow, and Simscape environments.
[DYNAMIC_EXECUTION]: The skill references the use of evaluate_matlab_code to execute MATLAB scripts for tasks that exceed the scope of standard editing tools, such as programmatically creating behavior models or defining complex Stateflow transitions. These operations use documented MATLAB and Stateflow APIs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted model data via model_read to understand existing topology before editing. This surface could potentially be used to influence agent behavior through maliciously named blocks or signal properties.
Ingestion points: model_read calls in SKILL.md, references/system-composer.md, and references/stateflow.md.
Boundary markers: The instructions recommend resolving paths from unique IDs (blk_X) rather than manually constructing names, which mitigates some risks, though explicit delimiters for model content are absent.
Capability inventory: model_edit (structural changes), evaluate_matlab_code (arbitrary MATLAB scripting).
Sanitization: No specific sanitization or filtering of block names or signal labels is described in the workflow instructions.

building-simulink-models