The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation encourages installing the belt-sh/cli via npx and provides links to installation scripts and guidelines hosted on a remote GitHub repository (inference-sh/skills). These external resources are required for the skill's intended functionality of performing web searches and extraction.\n- [PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted content from the web for processing by LLMs, creating an indirect prompt injection surface where malicious instructions in web pages could influence the agent's behavior.\n
Ingestion points: Web content retrieved through tavily/search-assistant, tavily/extract, exa/search, and exa/extract tools.\n
Boundary markers: The provided workflows do not include delimiters or specific instructions to the agent to disregard instructions embedded within the retrieved content.\n
Capability inventory: The skill utilizes the belt CLI to interact with search APIs and external language models (e.g., Claude via OpenRouter).\n
Sanitization: There is no evidence of content filtering, escaping, or validation before interpolating external search results into LLM prompts.

web-search