video-to-superprompt

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes command-line tools ffmpeg and ffprobe to process video files, extract frames, and retrieve technical metadata.\n
  • Evidence: SKILL.md includes shell commands for frame extraction (ffmpeg -y -i "$VIDEO" ...) and metadata inspection (ffprobe -v error ...).\n- [INDIRECT_PROMPT_INJECTION]: The skill operates on untrusted external data by ingesting video files from local paths and remote URLs, which represents a potential surface for indirect prompt injection if the processed content contains malicious instructions.\n
  • Ingestion points: Local video paths, uploaded files, and external URLs mentioned in SKILL.md.\n
  • Boundary markers: None specified to delimit ingested content from agent instructions.\n
  • Capability inventory: Access to shell execution for media tools and file system writes for frame storage.\n
  • Sanitization: No explicit content validation or sanitization is performed on the video metadata or content before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 01:30 AM
Security Audit — agent-trust-hub — video-to-superprompt