The Agent Skills Directory

[SAFE]: The skill is a legitimate documentation resource for the MATLAB Image Processing Toolbox. It uses the evaluate_matlab_code tool for intended functional verification of code snippets.
[INDIRECT_PROMPT_INJECTION]: The skill naturally interacts with external image data, which is standard for image processing tasks. While this presents an inherent surface area for indirect prompt injection (e.g., through malicious metadata in user-provided images), the skill does not contain any instructions that would facilitate an exploit.
Ingestion points: Files references/image-io.md, references/image-understanding.md, and references/deep-learning-imaging.md describe loading images, videos, and medical data from disk or folders.
Boundary markers: The instructions do not include specific boundary markers or warnings to disregard instructions potentially embedded in image data or metadata.
Capability inventory: The skill uses the evaluate_matlab_code tool to execute scripts. It also has capabilities for writing files (imwrite, VideoWriter, exrwrite) and reading system information via ver and hasIPT.
Sanitization: There are no documented sanitization steps for input data before it is processed or visualized.
[COMMAND_EXECUTION]: The SKILL.md file explicitly directs the agent to "Execute code blocks via evaluate_matlab_code before presenting them to the user." This core functionality allows the agent to verify code correctness and is used in a benign, documentation-focused context.
[EXTERNAL_DOWNLOADS]: The skill references several official MathWorks support packages (e.g., for Segment Anything Model and Circle Detection). These involve downloading pre-trained models and additional components from MathWorks servers, which are trusted resources within the MATLAB ecosystem.

matlab-process-images