image-to-image-character-generation

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches character reference images from external storage providers.\n
  • The skill retrieves images from supabase.co and googleapis.com to use as face anchors and pose references.\n- [INDIRECT_PROMPT_INJECTION]: The skill builds image generation prompts by combining character anchors with dynamic scene descriptions.\n
  • Ingestion points: Scene descriptions and attire/detail/mood components provided at runtime in SKILL.md.\n
  • Boundary markers: No specific delimiters or safety instructions are defined to wrap these components in the final prompt composition.\n
  • Capability inventory: Performs network API calls to api.kie.ai, writes files to the local workspace directory, and executes post-generation scripts.\n
  • Sanitization: No explicit sanitization or validation of the input scene components is performed before interpolation.\n- [COMMAND_EXECUTION]: The skill workflow involves executing external scripts for social media posting.\n
  • The documentation mentions that on approval, images are posted via platform-specific scripts, indicating a capability to launch subprocesses.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 09:40 PM
Security Audit — agent-trust-hub — image-to-image-character-generation