The Agent Skills Directory

[SAFE]: The skill integrates with the inference.sh CLI tool. Access is properly scoped in the YAML frontmatter using the allowed-tools field, restricting the agent's execution environment to the infsh command.
[SAFE]: All external references and download links, including the CLI installation instructions and app repositories, point to official vendor domains (inference.sh) or the vendor's GitHub organization (inference-sh). These are verified vendor resources according to the provided author context.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface, as it is designed to fetch content from the open web and process it with an LLM. While this is a common pattern for search skills, it introduces a potential vector for malicious content to influence the agent.
Ingestion points: Data enters the context via tavily/extract, exa/extract, and search result outputs.
Boundary markers: The examples use XML-like tags (e.g., <search-results>, <content>) to delineate external data from the system prompt.
Capability inventory: The tool allows the agent to run CLI applications for search and LLM processing.
Sanitization: No explicit sanitization of the retrieved web content is documented in the provided examples; the skill relies on the LLM's internal guardrails and the use of delimiters.

web-search