The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection, as it encourages the agent to ingest and process untrusted external data (filenames and paths) which are then used in executable MATLAB code.
Ingestion points: The agent ingests file names, directory paths, and project labels provided by the user or found in the filesystem (as described in SKILL.md and references/path-and-file-management.md).
Boundary markers: The instructions lack explicit boundary markers or "ignore embedded instructions" warnings for the user-supplied strings used in API calls.
Capability inventory: The skill requires the evaluate_matlab_code tool, which allows the execution of arbitrary MATLAB commands, including file system modifications and project property changes.
Sanitization: There is no mention of sanitizing, escaping, or validating user-provided file names or paths before they are interpolated into MATLAB commands like addFile, addPath, or set_param.

managing-simulink-projects