Local web search using SearXNG

**Quick judgment:** A solid, privacy-friendly general web search skill that’s very useful whenever you need current, web-based information, assuming the user has a working SearXNG instance configured. **What it does well:** - Routes generic “search the web / google / look up / research” requests to a self-hosted SearXNG, aggregating multiple engines without API keys. - Returns clean, structured JSON (title, URL, description, engines, score), making it easy to plug into other tools or chain in workflows. - Supports basic tuning via result count and language flags, including simple news/region bias via query wording and --lang. **Key risks / limitations:** - Fully depends on the user’s SearXNG instance (SEARXNG_URL) being correctly set up, reachable, and well-configured; misconfiguration or downtime will break the skill. - Search quality and freshness vary with SearXNG’s upstream engines and their current behavior, which you do not fully control. - Like any generic web search, it can surface unsafe, biased, or low‑quality content; additional filtering or validation may be needed in sensitive workflows. - Does not expose advanced search operators or rich filters beyond count and language; deep vertical/search customization must be handled at the SearXNG level. **Recommended scenarios:** - Use as the default tool when a user explicitly asks to “search the web / google / look up / research” something, especially if timeliness matters. - Employ in research or investigative workflows where combining multiple search engines is beneficial and the user is privacy-conscious. - Avoid as the sole source of truth for high‑stakes decisions (medical, legal, financial, safety‑critical); pair with domain-specific tools or verified sources and apply additional reasoning and validation.