The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute belt CLI commands for platform authentication (belt login) and running remote AI inference applications. This is the primary mechanism for the skill's functionality.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by ingesting and processing external data during its workflows.
Ingestion points: The image_url parameter within the input JSON of belt app run commands in SKILL.md (e.g., https://your-image.jpg).
Boundary markers: Absent; the skill instructions show direct interpolation of URLs into the command arguments without specific delimiters or guards against malicious content within the external resource.
Capability inventory: Shell command execution via the Bash tool, including network-enabled CLI tools.
Sanitization: The instructions do not include validation or sanitization steps for the external URLs or the resulting output from the inference apps before further processing.

image-upscaling