The Agent Skills Directory

[OBFUSCATION]: The file scripts/extractDiagnostics.p is a MATLAB P-code file, which is an opaque binary format that prevents static security analysis of the logic used to extract diagnostics from profiling reports.
[DYNAMIC_EXECUTION]: The skill dynamically executes generated MEX binaries using the feval function in scripts/benchmarkMex.m to perform performance benchmarking and numerical verification.
[COMMAND_EXECUTION]: The skill invokes the codegen command to compile user-provided MATLAB design files into CUDA-based MEX or library targets, which involves executing a build process on potentially untrusted source code.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes user-supplied MATLAB files and data, creating a surface for indirect prompt injection where malicious instructions could attempt to influence the agent's optimization or diagnostic-fix logic.
Ingestion points: User-provided MATLAB design files and input data resolved in SKILL.md (Step 1c).
Boundary markers: The skill does not use explicit boundary markers or delimiters when processing user-provided code, relying instead on numerical equivalence checks against the original file.
Capability inventory: The skill uses codegen for compilation, benchmarkMex for execution, and gpuPerformanceAnalyzer for profiling.
Sanitization: No sanitization or validation is applied to the user-provided code or input values before they are passed to the compilation and execution tools.

matlab-optimize-gpu-codegen