magic-hour-image-editing
Pass
Audited by Gen Agent Trust Hub on Nov 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references API schemas, documentation, and implementation guides located on vendor-owned domains, including
docs.magichour.ai,mcp.magichour.ai, andgithub.com/magichourhq. These are standard technical references for the skill's operation. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to interpret and act upon user-provided images and text-based editing requests.
- Ingestion points: User-attached images and natural language requests described in
SKILL.md. - Boundary markers: Instructions explicitly require the agent to establish an "edit boundary," separating elements to be changed from those to be preserved (identity, geometry, pose).
- Capability inventory: The skill utilizes image editing tools and authenticated API endpoints (
POST /v1/ai-image-editor) to fulfill requests. - Sanitization: The instructions provide a specific prompt structure designed to isolate changes and protect sensitive image details from being altered by ambiguous user input.
Audit Metadata