og-image-design
Warn
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
belt-sh/clipackage and references installation guidelines hosted athttps://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md. These resources originate from an external third-party source not currently listed as a verified vendor. - [REMOTE_CODE_EXECUTION]: The core functionality of the skill involves using the
beltCLI to invoke and run remote applications (such asinfsh/html-to-imageandfalai/flux-dev-lora) hosted on the inference.sh infrastructure. - [COMMAND_EXECUTION]: The skill provides commands for shell execution, including logging into an external service (
belt login) and running specialized tools (belt app run). - [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing untrusted data by taking HTML strings and user-defined prompts as input for the underlying image generation tools.
- Ingestion points: HTML input strings in
belt app runcommands and prompts for AI image generation. - Boundary markers: Absent. The HTML and prompt content are passed directly within JSON strings to the CLI.
- Capability inventory: Execution of shell commands via the
beltCLI and network access to remote AI services. - Sanitization: None described in the skill instructions; safety relies entirely on the underlying remote applications.
Audit Metadata