Tavily

**Judgement:** A solid, focused web-search skill that leverages Tavily to give OpenClaw agents fresh, source-rich web results. It’s best suited for scenarios where up-to-date URLs and references matter more than deep, built-in analysis. **What it does well:** - Provides a straightforward way to run Tavily searches (via skill + CLI) and get JSON results and URL lists. - Supports practical filters (max results, include/exclude domains) to target authoritative sources. - Integrates cleanly into scripts and automation thanks to structured JSON and simple CLI usage. **Key risks & limitations:** - **External dependency:** Requires a Tavily API key and stable access to Tavily’s service; outages, rate limits, or API changes will affect reliability. - **Data privacy:** All queries and retrieved URLs go to an external third-party service; not suitable for highly sensitive or regulated data without careful review. - **Partial API coverage:** The bundled CLI only exposes a subset of Tavily’s fields; more advanced Tavily features may not be reachable through this skill. - **Content safety:** The skill does not appear to add its own content filtering; unsafe or low-quality pages can still appear in results depending on the query. **Recommended scenarios:** - Keeping an agent’s answers current by pulling recent links and references from the web. - Technical or research-oriented workflows where you want controlled domain filters (e.g., only standards bodies or official docs). - Automated link-gathering or lead-research pipelines that consume JSON search results and then perform downstream analysis. **Less suitable for:** - Offline or strictly local-only deployments. - Workflows that require guaranteed filtering of harmful content or tight control over data egress. - Use cases needing deep interpretation of results inside the skill itself rather than just high-quality retrieval.