ai-image-generation

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references installation scripts and additional modular skills hosted on the vendor's official GitHub repository (inference-sh/skills).
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of user-provided text prompts for image generation, which represents a common surface for indirect injection attacks.
  • Ingestion points: User input is interpolated into the --input argument of the belt app run command within the SKILL.md file examples.
  • Boundary markers: The skill utilizes structured JSON objects (e.g., '{"prompt": "..."}') to encapsulate user input, providing a clear boundary between instructions and data.
  • Capability inventory: The skill's execution environment is strictly limited to the vendor's command-line interface as defined by the allowed-tools: Bash(belt *) restriction.
  • Sanitization: The skill does not implement explicit client-side sanitization of prompt strings, relying instead on the safety filters of the target AI models (e.g., FLUX, Gemini, OpenAI).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 08:27 PM
Security Audit — agent-trust-hub — ai-image-generation