The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of media processing tools such as ffmpeg, ffprobe, magick, and vips. All provided script examples demonstrate secure execution patterns, such as proper shell variable quoting and list-based parameter passing in Python, which effectively mitigate command injection risks.- [DYNAMIC_EXECUTION]: The skill includes a Python script (make_thumbnails.py) that utilizes the subprocess module to invoke external binaries. The implementation uses best practices for dynamic execution by passing arguments as a list rather than a shell string, ensuring that user-provided file paths do not lead to arbitrary code execution.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted media files from external directories, representing a potential injection surface for malicious filenames or metadata. However, this risk is thoroughly addressed by the skill's design principles. 1) Ingestion points: Files are read from directories specified by IN_DIR and command-line arguments. 2) Boundary markers: The instructions mandate re-probing outputs and manual visual inspections (Look field) to verify integrity. 3) Capability inventory: Binary execution is limited to specific, well-defined media processing tools. 4) Sanitization: All logic uses secure parameter handling for system calls.- [SAFE]: No obfuscation, persistence mechanisms, or unauthorized network operations were detected. All external references in the skill metadata and documentation are for attribution to reputable open-source projects.

media-processing