mage-video-editor

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external video content and user-provided text instructions which are interpolated into templates for downstream generation tools.
  • Ingestion points: Video URLs, reference images, and user instructions ingested from chat or local file uploads via the create_upload tool.
  • Boundary markers: The skill employs a comprehensive "wrapped prompt" for Cherry models, specifically instructing the AI to preserve the source video's identity and only apply minimal requested changes.
  • Capability inventory: Uses ffprobe for metadata extraction, create_upload for file handling, and a remote generate tool for video processing.
  • Sanitization: Relies on the integrated safety filters of the Mage service (as indicated by the content_blocked error handling).
  • [SAFE]: All external resources and API tools trace back to the official vendor infrastructure (mage-space).
  • [SAFE]: The use of ffprobe for video duration measurement is a standard, non-destructive operation for calculating processing costs.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 08:03 PM