find-movie-clips
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill instructions and associated script focus on legitimate media processing workflows for generating AI reference material.
- [COMMAND_EXECUTION]: Uses
ffmpeg,ffprobe, and a local bash script (scripts/takes.sh) to segment video files, detect scene cuts, and generate contact sheets. This command execution is appropriate for the skill's stated purpose of video editing. - [EXTERNAL_DOWNLOADS]: Fetches video files, metadata, and subtitles from trusted and well-known services, specifically YouTube and the Internet Archive, using
yt-dlpandcurl. All network operations target established media repositories. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content such as subtitles (VTT/SRT) and metadata from media platforms. While this presents an ingestion surface for potentially malicious instructions, the skill uses sanitization tools like
jqandgrepfor targeted extraction, and the instructions emphasize human review of the final output files. - Ingestion points: YouTube subtitles via
yt-dlp, Internet Archive metadata viacurlandjq. - Boundary markers: None explicitly for text processing, though human review of visual 'contact sheets' is mandated as a final check.
- Capability inventory: Subprocess calls to
ffmpeg,yt-dlp,curl, and execution of the includedscripts/takes.shfile. - Sanitization: Employs
jqfor structured JSON parsing andgrepfor filtering specific subtitle content.
Audit Metadata