video-ocr

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes ffmpeg as a subprocess to handle frame extraction from input video files.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of ffmpeg via Homebrew and several Python libraries (pyobjc-core, pyobjc-framework-Vision, pyobjc-framework-Cocoa, pillow) from standard package registries.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted text content extracted from video frames (e.g., presentations, tutorials, or screencasts).
  • Ingestion points: Video file content processed via the video_ocr.py script.
  • Boundary markers: None specified in the documentation to delineate OCR results from agent instructions.
  • Capability inventory: File system writes (JSONL, Markdown, JPG frames) and subprocess execution of ffmpeg.
  • Sanitization: No sanitization or filtering of the extracted text is mentioned before it is presented to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 09:40 PM
Security Audit — agent-trust-hub — video-ocr