gpt-image
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references and retrieves configuration and additional skills from the inference.sh GitHub organization.\n- [COMMAND_EXECUTION]: The skill utilizes the
beltcommand-line utility to perform image generation and editing tasks.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data including text prompts, reference image URLs, and mask URLs (ingestion points in SKILL.md) and passes them to thebeltCLI tool (capability in SKILL.md). While there are no explicit boundary markers or sanitization logic defined in the skill instructions, this represents a standard operating surface for image generation tools and relies on the safety filters of the underlying GPT-Image-2 model.
Audit Metadata