fountain-clip-producer

Pass

Audited by Gen Agent Trust Hub on Oct 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on system binaries including ffmpeg, ffprobe, yt-dlp, magick (ImageMagick), and fc-match (fontconfig) to process media files. These are invoked using Python's subprocess.run with argument lists, which is the recommended method for executing external commands without invoking a shell.
  • [EXTERNAL_DOWNLOADS]: The skill fetches video and audio content from external sources like YouTube and the Fountain API via yt-dlp. It also suggests downloading a 141 MB whisper.cpp model from a well-known repository on Hugging Face (huggingface.co/ggerganov/whisper.cpp) to enable word-level transcription during the preflight stage.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as episode transcripts and auto-generated captions. 1. Ingestion points: External transcripts and captions are read into the skill context via modules/media/scripts/build-time-map.py and modules/captions/scripts/build-captions.py. 2. Boundary markers: No specific delimiters or 'ignore' instructions were found for the interpolated text data. 3. Capability inventory: The skill uses subprocess.run to call ffmpeg, magick, and yt-dlp across several modules (media, captions, overlays, framing). 4. Sanitization: The modules/overlays/scripts/build-overlays.py script includes an escape_drawtext function designed to escape special characters before they are passed to FFmpeg filters, and build-captions.py performs text hygiene (faithful-clean) on word data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 29, 2026, 10:26 AM