The Agent Skills Directory

[SAFE]: The skill implements a legitimate workflow for media production, focusing on creative direction and project maintenance. It includes specific commands for user validation and cost ceilings, demonstrating safe operational practices.
[INDIRECT_PROMPT_INJECTION]: The skill processes external media files from the uploads/ directory, which presents an attack surface for indirect prompt injection if those files contain malicious metadata or instructions. 1. Ingestion points: User-owned source material in uploads/ directory (SKILL.md). 2. Boundary markers: Mentions treating uploads as immutable source material but lacks explicit instruction delimiting for ingested content. 3. Capability inventory: Executes local media utilities (file, ffprobe), writes artifacts to project-specific paths, and makes calls to external generation providers. 4. Sanitization: Implicitly relies on standard media parsing libraries; no specific text-level sanitization is defined in the instructions.

video-creation