p-image
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references installation instructions and documentation hosted on GitHub (
raw.githubusercontent.com/inference-sh/) and the platform's official site (inference.sh). These are well-known or platform-specific sources intended for skill setup. - [COMMAND_EXECUTION]: The skill utilizes the
beltCLI tool (restricted viaBash(belt *)) to interact with remote image generation applications. Commands likebelt loginandbelt app runare standard for the platform's operation. - [DATA_EXFILTRATION]: User-provided text prompts and image URLs are sent to the
inference.shplatform for processing. This is the core functionality of the skill and is documented as the intended data flow. - [INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by accepting external data such as image URLs and user-defined prompts. These inputs are interpolated into CLI commands (e.g.,
--input '{"prompt": "..."}'). While no specific sanitization is mentioned in the markdown, the use of JSON-structured input for the CLI provides a basic boundary.
Audit Metadata