Privacy-first web search with DuckDuckGo-style bangs (!w, !yt, !gh)

**Judgement:** This skill is a solid, privacy-first web/metasearch tool with powerful DuckDuckGo-style bang support and good JSON output, well-suited as a primary search capability for agents that need privacy and multi-engine coverage. **Strengths** - Strong privacy posture: IP masking, no cookies/referrers, randomized browser fingerprints, optional Tor/proxy. - Aggregation from 250+ engines and multiple categories (general, news, images, videos, science). - Bangs (!w, !yt, !gh, !r, etc.) enable precise, engine-specific queries from a single interface. - Returns clean JSON (title, URL, snippet) that is easy to consume in agent workflows. - Flexible configuration: can use a self-hosted SearXNG instance or public instances via `SEARXNG_URL`. **Key Risks / Limitations** - **Dependency on a running SearXNG instance:** By default expects `http://127.0.0.1:8080`; without a properly configured local or public instance, the tool will fail. - **HTML parsing fragility:** It scrapes HTML (since JSON API is often disabled); changes in SearXNG or upstream engine HTML can break parsing and reduce reliability. - **Public instance variability:** Using community instances can introduce rate limits, downtime, or slower performance; reliability is lower than a dedicated, self-hosted setup. - **Content safety:** Like any generic web search, results can include harmful, NSFW, or misleading content. The description doesn’t mention safety filters, so agents using this should layer their own content evaluation/filters. - **Potential ToS/legal considerations:** SearXNG’s aggregation/scraping model may be in tension with some search engine terms of service; responsibility lies with the deployer. **Recommended Scenarios** - Primary search tool for privacy-sensitive workflows where tracking, profiling, or commercial APIs are undesirable. - High-volume or exploratory web research where commercial search APIs might be too costly or rate-limited. - Tasks needing multi-engine or multi-category coverage (general + news + science) with a single query. - Engine-specific searches via bangs (Wikipedia, YouTube, GitHub, Reddit, etc.) when an agent needs targeted results without bespoke integrations. - Environments with self-hosted SearXNG infrastructure, where reliability and speed can be controlled. **Less Suitable For** - Contexts requiring guaranteed uptime, strict SLAs, or highly stable schemas (due to HTML parsing and reliance on external instances). - Workflows that demand strong built-in content filtering or safety guarantees without additional layers. - Deployments where maintaining a SearXNG instance or relying on public instances is not acceptable.