The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted external sources (the public web) via search and crawling endpoints, which is a standard surface for indirect prompt injection attacks.
Ingestion points: The Tavily API's search, extract, crawl, and research methods retrieve content from arbitrary URLs across the internet, as documented in references/search.md, references/extract.md, references/crawl.md, and references/research.md.
Boundary markers: The API returns content in specific fields like raw_content or results, providing a structural boundary for the agent to distinguish between its instructions and retrieved data.
Capability inventory: The skill is intended to supply data to agents that may perform various tasks, including synthesizing research and executing tool calls based on search results.
Sanitization: While the documentation does not detail server-side sanitization of the scraped content, it follows best practices by providing structured responses for agent processing.
[CREDENTIALS_UNSAFE]: The skill uses safe practices for secret management by recommending the use of environment variables (e.g., TAVILY_API_KEY) and providing standard placeholders like tvly-YOUR_API_KEY for demonstration purposes.

tavily-best-practices