The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references documentation and installation guides hosted on the official GitHub repository for inference-sh. These resources are directly related to the skill's primary function of interfacing with the platform.
[COMMAND_EXECUTION]: The skill executes belt CLI commands through the Bash tool. The allowed-tools frontmatter field restricts the execution scope to only commands starting with belt, adhering to the principle of least privilege.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface for processing user-supplied image prompts and URLs.
Ingestion points: The --input JSON payload in belt app run commands (SKILL.md).
Boundary markers: User input is encapsulated within structured JSON strings to separate it from command-line arguments.
Capability inventory: Command execution is strictly limited to the belt application namespace.
Sanitization: The skill relies on the inference platform's underlying model guardrails for input safety.

p-image