The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the belt CLI tool. This execution is properly scoped and restricted via the allowed-tools frontmatter configuration.
[EXTERNAL_DOWNLOADS]: The documentation references installation scripts and secondary skills located in the inference-sh organization's public GitHub repositories. These resources are necessary for the intended functionality of the platform.
[PROMPT_INJECTION]: The skill operates by taking user input and passing it to external language models, creating a surface for indirect prompt injection if malicious instructions are embedded in the data being processed.
Ingestion points: The prompt and system JSON fields provided to the belt app run command in SKILL.md.
Boundary markers: No explicit delimiter or protection instructions are provided in the command templates.
Capability inventory: The belt tool allows for authenticated LLM inference and interaction with various platform-hosted applications.
Sanitization: Input sanitization and safety filtering are deferred to the underlying LLM provider (e.g., Anthropic, Google).

llm-models