magic-hour-brand-kit
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill connects to the vendor's official API and documentation sites to retrieve assets and integration guides.
- Evidence: References to
https://mcp.magichour.ai/andhttps://docs.magichour.ai/api-referencefor platform interaction. - Evidence: Downloads project-specific brand assets from official Magic Hour URLs generated during the branding process.
- [INDIRECT_PROMPT_INJECTION]: The skill serves as a processing engine for user-supplied brand assets, creating a potential surface for indirect instructions.
- Ingestion points: Processes brand briefs, logos, and documentation provided by the user or fetched via vendor tools (
SKILL.md). - Boundary markers: Relies on natural language context rather than strict technical delimiters to separate brand content from system instructions.
- Capability inventory: Includes capabilities for generating images, polling project status, and writing local documentation files (
SKILL.md,references/applications.md). - Sanitization: Emphasizes manual inspection of visual output and dimensions by the agent to ensure fidelity and compliance with the brand guide.
Audit Metadata