gemini-websearch

Web searches with caching store results by MD5-keyed query hash with a 1-hour TTL to avoid repeated requests, while validation scoring filters low-quality results. This Skill implements both via Gemini CLI's google_web_search tool in headless mode, returning JSON-formatted validated results.

Yes. This Skill provides built-in analytics that track latency, cache performance metrics, and query quality scores. Analytics help identify bottlenecks and measure the efficiency gains from caching across single and batch search operations.

Batch mode processes multiple queries in a single execution with optional retry on failure. The Skill accepts batch inputs, applies the same validation and caching logic to each query, and outputs aggregated JSON results with per-query analytics.

Search results return in JSON format with structured fields including validated content, relevance scores, latency metrics, and cache status. This format enables downstream processing, filtering, and integration into analytics pipelines.

Yes. The Skill handles both use cases—searching for current documentation, release notes, and API references, as well as general user-requested web queries. The same caching, validation, and analytics apply across all query types.

Repeated queries within the 1-hour TTL return cached results instantly without hitting the web search tool, reducing latency and API calls. Cache misses trigger a fresh search, update the cache, and reset the TTL counter.