The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external image URLs which are passed as arguments to the belt CLI tool. This creates a surface where malicious instructions could be embedded in the metadata or data of the processed images.
Ingestion points: The image_url parameter within the --input JSON block used by the belt app run command in SKILL.md.
Boundary markers: None provided to delimit the untrusted URL input.
Capability inventory: The skill is restricted to executing the belt CLI tool via the Bash tool.
Sanitization: No explicit sanitization or validation of the external content is defined in the instructions.
[EXTERNAL_DOWNLOADS]: The skill references external installation documentation for its required CLI tool.
Evidence: Link to https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md for CLI installation instructions.

image-upscaling