background-removal
Pass
Audited by Gen Agent Trust Hub on Nov 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references installation scripts and documentation hosted on the official GitHub repository for the inference.sh platform (e.g.,
https://raw.githubusercontent.com/inference-sh/skills/...). These are standard resources for the intended functionality. - [COMMAND_EXECUTION]: The skill uses the
beltCLI tool to execute remote applications (e.g.,infsh/birefnet,falai/reve) for processing images. These commands are scoped to the platform's legitimate application ecosystem. - [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect injection as it accepts external
image_urlinputs to be processed by remote AI models. - Ingestion points: The
image_urlfield in the JSON input forbelt app runcommands (SKILL.md). - Boundary markers: None present in the provided command examples.
- Capability inventory: Uses the
Bashtool to run thebeltCLI for remote inference. - Sanitization: Sanitization is handled server-side by the inference platform; the skill itself does not perform local validation of the image content.
Audit Metadata