The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the belt CLI utility using npx and references official documentation and configurations hosted on the inference.sh GitHub repository.
[COMMAND_EXECUTION]: The skill utilizes the belt command-line tool via the Bash tool to interact with remote AI image generation applications such as those provided by ByteDance and Fal.ai.
[DATA_EXFILTRATION]: Contains instructions for belt login, which handles authentication for the underlying service. No unauthorized access to local sensitive files or non-vendor domains was detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined prompts within structured JSON inputs for image generation. While this presents a standard attack surface for prompt injection, the risk is minimized by the use of JSON boundaries and the nature of the image generation tasks.
Ingestion points: Prompt fields within the --input argument of belt app run commands (SKILL.md).
Boundary markers: Uses JSON object structures to encapsulate external prompt content.
Capability inventory: Shell execution restricted to the belt command via allowed-tools configuration.
Sanitization: Relies on the underlying CLI and API endpoints for sanitization of prompt data.

product-photography