flux-image

Pass

Audited by Gen Agent Trust Hub on Jul 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and fetches configuration or documentation from the official inference.sh GitHub repository.
  • [COMMAND_EXECUTION]: Executes the 'belt' CLI tool to run remote inference applications. This execution is restricted to the 'belt' command via the 'allowed-tools' frontmatter, adhering to the principle of least privilege.
  • [PROMPT_INJECTION]: The skill contains an indirect prompt injection attack surface because it processes untrusted external data.
  • Ingestion points: Processes user-provided text prompts and external image URLs ('image_url') through shell command arguments in SKILL.md.
  • Boundary markers: Input parameters are encapsulated within JSON structures (e.g., --input '{"prompt": "..."}'), which provides basic delimitation.
  • Capability inventory: The skill can execute the 'belt' command to perform remote image generation and manipulation tasks.
  • Sanitization: No explicit sanitization or escaping of the user-provided prompt or URL strings is performed within the skill instructions before they are interpolated into the JSON payload.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 22, 2026, 07:31 AM
Security Audit — agent-trust-hub — flux-image