The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the belt CLI tool to interact with the inference.sh platform. The allowed-tools frontmatter correctly restricts the agent's environment to this specific command-line interface, following the principle of least privilege.
[EXTERNAL_DOWNLOADS]: The skill references official documentation and installation scripts from the inference.sh domain and its associated GitHub repository (inference-sh/skills). These are recognized as legitimate resources for the skill's primary function and do not represent a security risk.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text prompts and external image URIs, which are passed as JSON input to the belt command.
Ingestion points: The prompt and reference_images parameters documented in SKILL.md.
Boundary markers: None explicitly defined in the provided examples; however, the input is structured as JSON which provides a natural data/instruction boundary.
Capability inventory: The skill is limited to executing the belt CLI tool via the Bash tool.
Sanitization: The skill relies on the underlying belt tool and the inference.sh API to handle and sanitize the provided input.

qwen-image-2