skills/inference-sh/skills/flux-image/Gen Agent Trust Hub

flux-image

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 instructions from the vendor's official GitHub repository (inference-sh/skills) and suggests installing a related CLI skill (belt-sh/cli). These resources are part of the vendor's ecosystem and are necessary for the skill's intended functionality.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides an interface for generating images based on user-provided prompts and URLs, creating an indirect prompt injection surface.\n
  • Ingestion points: User prompts and image URLs are passed into the belt app run command via the --input parameter in SKILL.md.\n
  • Boundary markers: The skill uses JSON formatting for input parameters, which helps distinguish user-provided data from command-line arguments.\n
  • Capability inventory: The skill utilizes shell command execution through the belt CLI tool (authorized via allowed-tools: Bash(belt *)) to communicate with remote image generation models.\n
  • Sanitization: The skill does not implement custom sanitization for prompts, relying on the safety guardrails of the underlying FLUX models and the inference.sh platform.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 09:30 PM
Security Audit — agent-trust-hub — flux-image