The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and examples demonstrate the use of the Bash tool to execute the tvly CLI for search and extraction, alongside python3 and jq for filtering JSON data locally.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the public web via search results, creating a potential surface for indirect prompt injection. The skill provides strategies to mitigate this by filtering data locally before returning it to the agent's context.
Ingestion points: Data retrieved via tvly search and tvly extract commands (found in SKILL.md).
Boundary markers: The skill instructs the agent to return only necessary evidence, though it does not provide explicit prompt delimiters in the code snippets.
Capability inventory: The skill utilizes file writing (via CLI -o flag) and reading (via Python Path.read_text()), as well as general shell execution via the allowed Bash tools.
Sanitization: Python snippets included in the skill show field-specific extraction (e.g., title, url) and string slicing (e.g., [:300], [:1200]) to limit the ingestion of raw, untrusted content.

tavily-dynamic-search