The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it fetches and processes content from external, untrusted websites.
Ingestion points: Untrusted data enters the agent context via the Tavily and Exa search/extract apps (e.g., tavily/extract, exa/search) as described in SKILL.md.
Boundary markers: The skill instructions do not recommend using delimiters or specific 'ignore instructions' markers when interpolating search results into LLM prompts.
Capability inventory: The agent uses Bash(infsh *) to run extraction and search tasks via a command-line interface.
Sanitization: No sanitization or filtering of the retrieved web content is documented before it is passed to other tools or LLMs.
[COMMAND_EXECUTION]: The skill requires the use of the infsh CLI to perform its primary functions, involving shell command execution to interface with the inference.sh platform.
[EXTERNAL_DOWNLOADS]: The documentation points to external installation scripts for the CLI tool and suggests installing additional skill components from the inference-sh GitHub organization via the npx package runner.

web-search