ugc-character-consistency

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill uses a template that interpolates user-provided action descriptions into video generation prompts.
  • Ingestion points: The action variable in the clip() function (SKILL.md) accepts arbitrary user input to describe video sequences.
  • Boundary markers: User input is directly combined with the LOOK string using f-string interpolation (f"{action} {LOOK} ...") without delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill utilizes generate_video, create_brand_asset, and generate_brand_asset tools which handle visual identity and content creation.
  • Sanitization: No validation, escaping, or filtering is performed on the action input before it is sent to the model.
  • [EXTERNAL_DOWNLOADS]: The skill references an external Model Context Protocol (MCP) server for tool integration.
  • Evidence: The setup instructions include the command claude mcp add --transport http apimage https://mcp.apimage.org/mcp.
  • Context: The domain apimage.org belongs to the author's official infrastructure for providing brand asset and video services.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 09:15 AM