The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation provides numerous examples of using the belt CLI tool via Bash to execute AI model applications. This includes tasks like generating images (pruna/p-image), creating videos (pruna/p-video-avatar), and merging media assets (infsh/media-merger).
[EXTERNAL_DOWNLOADS]: The skill references external URLs for installation instructions hosted on GitHub and utilizes several external URLs for sample images, audio files, and videos throughout the production workflow guides.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface because it processes untrusted data from external media URLs.
Ingestion points: Untrusted data enters the agent's context through URL parameters such as image, audio, and video_url within the belt app run commands documented in SKILL.md.
Boundary markers: There are no explicit markers or instructions to ignore potential instructions embedded in the processed media assets.
Capability inventory: The skill possesses the capability to execute shell commands and run AI applications via the belt CLI tool.
Sanitization: There is no evidence of sanitization, validation, or filtering of the provided URLs or the content they retrieve before processing.

talking-head-production