The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions to install the belt CLI tool and other platform-related skills using npx skills add commands targeting the belt-sh and inference-sh namespaces. It also references installation scripts hosted on raw.githubusercontent.com.
[COMMAND_EXECUTION]: The skill requires the Bash tool and specifically requests authorization to execute the belt command via allowed-tools: Bash(belt *). This follows the principle of least privilege by restricting shell access to a specific utility.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for user-provided data to be processed by external Large Language Models (LLMs).
Ingestion points: User-supplied strings are interpolated into JSON payloads passed to the belt app run command in SKILL.md.
Boundary markers: The skill uses JSON structured objects (e.g., {"prompt": "..."}) to encapsulate user input, providing a basic structural boundary.
Capability inventory: The skill utilizes the Bash tool to execute the belt CLI tool, which performs network operations to communicate with model inference servers.
Sanitization: No explicit client-side sanitization or filtering of the prompt content is described in the skill instructions; it relies on the safety filters of the destination LLMs.

llm-models