video-analyser
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill extracts text via OCR and transcribes audio from external video files, providing this data directly to the AI model.
- Ingestion points: Processes frames and audio from video files located at local paths, provided URLs, or Linear issue attachments (SKILL.md, Step 6 and Step 7).
- Boundary markers: The skill uses delimited blocks like <ocr_frame_N> and <audio_transcript> to separate extracted content from instructions (SKILL.md, Step 8).
- Capability inventory: The skill has access to shell execution for media processing and network access via curl (SKILL.md, Step 1 and Step 5).
- Sanitization: Content extracted from videos (text and audio) is passed to the analysis prompt without escaping or validation.
- [EXTERNAL_DOWNLOADS]: The skill downloads video files from user-supplied URLs and Linear ticket attachments using curl (SKILL.md, Step 1).
- [COMMAND_EXECUTION]: The skill makes extensive use of system commands for media analysis and processing, including ffmpeg, ffprobe, tesseract, and whisper (SKILL.md, Step 2, 5, 6, and 7).
Audit Metadata