The Agent Skills Directory

[SAFE]: The skill serves as a documentation and interface wrapper for the belt CLI tool, which is the official command-line interface for the inference.sh platform. All external references, including documentation links and installation instructions, point to the vendor's own verified domains and repositories (inference.sh and github.com/inference-sh).
[COMMAND_EXECUTION]: The skill is restricted to using the Bash(belt *) tool via its YAML frontmatter. This correctly implements the principle of least privilege by scoping the agent's capabilities to only the specific vendor CLI tool required for the task.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of image URLs and user-provided prompts. While this represents a standard attack surface for vision-capable models, the skill does not exhibit any specific vulnerabilities. The ingestion points include the --input parameters for prompts and image references. No specific boundary markers or sanitization logic are defined within the markdown itself, as these are typically handled by the underlying API and the belt tool.

gpt-image