The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references documentation and installation scripts hosted on the official GitHub repository for inference.sh.
[COMMAND_EXECUTION]: The skill utilizes the belt CLI tool to perform searches and content extraction. The execution environment is restricted to the belt command via the allowed-tools frontmatter configuration.
[PROMPT_INJECTION]: The skill processes untrusted content from the web and interpolates it into LLM prompts, which is a vector for indirect prompt injection.
Ingestion points: Data retrieved from tavily/search-assistant, tavily/extract, and exa/extract (SKILL.md).
Boundary markers: The prompt examples use tags like <search-results> and <content> to separate external data from instructions.
Capability inventory: Execution of the belt command through the Bash tool (SKILL.md).
Sanitization: The provided examples do not include explicit logic to sanitize or filter potential instructions within the retrieved web content.

web-search