Web Search by Desearch

**Verdict:** A solid, focused web search skill that provides structured SERP-style results suitable for most general web queries, with clear pagination and error handling. It is well-suited as a general-purpose search backend when you need fresh links and snippets beyond the model’s training data. **Strengths** - Real-time access to the web, giving up-to-date information. - Clean, predictable JSON structure (title, snippet, link) that is easy for agents to consume. - Simple pagination via a `--start` offset for multi-page search. - Explicit, machine-readable errors for invalid API keys and depleted balance. **Risks / Limitations** - Depends on an external Desearch API key, so outages, quota limits, or billing issues can break functionality. - No documented content filtering or safety layer on top of raw web results; agents may surface harmful, misleading, or NSFW content from the open web and must handle that themselves. - Only basic search is described (no advanced filters, language/region settings, etc.), which may limit precision for specialized queries. **Recommended Scenarios** - Use when you need current links and short descriptions (news, product info, documentation, recent discussions). - Use as a generic fallback web search tool for open-ended queries where other domain-specific tools are unavailable. - Avoid relying on it as the sole source for safety-sensitive or high-stakes decisions; combine with internal validation, cross-checking, and content moderation where needed.