The Agent Skills Directory

[OBFUSCATION]: The validation script scripts/+simscape/+agentic/sscIssues.p is provided as a MATLAB P-code file. P-code is an obfuscated and encrypted binary representation of MATLAB source code, which hides implementation details and prevents static analysis of the validation logic for hidden behaviors.
[DYNAMIC_EXECUTION]: The skill instructs the agent to execute the obfuscated P-code file using the evaluate_matlab_code tool. This involves loading and running pre-compiled binary logic that cannot be verified by the agent or the platform.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted Simscape source code provided by users to perform validation. This results in a vulnerability surface where the validation tool's output could be manipulated to inject instructions into the agent's decision-making process.
Ingestion points: User-supplied .ssc files containing component, domain, or function definitions.
Boundary markers: Not present. The workflow lacks specific instructions for the agent to treat tool output as data only or to ignore any instructions potentially embedded within it.
Capability inventory: The agent uses evaluate_matlab_code to execute validation logic and has the ability to read and write files within the project directory to implement fixes.
Sanitization: Sanitization of inputs and outputs cannot be verified due to the obfuscated nature of the primary validation tool (sscIssues.p).

simscape-write-ssc