The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the belt-sh/cli package via npx and fetches installation instructions from a remote GitHub repository (inference-sh/skills). These are standard operational requirements for utilizing the inference.sh platform.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text prompts and image URLs for image generation and editing tasks. While this introduces an ingestion surface for external content, the risk is inherent to the primary function of an AI image generation tool. 1. Ingestion points: User-provided prompt strings and images URL arrays in the belt app run command examples in SKILL.md. 2. Boundary markers: None explicitly defined in the instructions to separate user input from model instructions. 3. Capability inventory: Execution of the belt CLI tool via Bash(belt *) for remote API interaction. 4. Sanitization: Not explicitly documented; the skill relies on the downstream API's safety filters and the restricted scope of the belt command.
[SAFE]: No malicious patterns, such as command injection, data exfiltration, or obfuscation, were detected. The skill's operations are transparent and consistent with its described functionality.

nano-banana