What problem does it solve? Wiring together retrieval, generation, memory, and tool integrations for a research agent requires coordinating many moving parts across packages, and misconfigurations (wrong engine names, lost memory, shadowed exports) fail silently. This Skill guides you through the chat-agent-toolkit package so agents, search handlers, and persistence behave as intended. ## Core Features & Use Cases - Search-and-answer handlers: Configure webSearch, academicSearch, youtubeSearch, redditSearch, wolframAlphaSearch, and writingAssistant via createSearchHandlers with injectable retrieval functions. - Mastra agents, workflows, RAG, and evals: Build tool-using agents, multi-step research/RAG workflows, and score generations with factuality, relevance, coherence, and toxicity evals. - Memory and MCP tooling: Persist long-term memory with SimpleMemory, Drizzle, D1, or KV storage, and connect QwkSearch or third-party OAuth MCP tool sessions. - Use Case: You are adding a new tool to a research agent on Cloudflare Workers — use this Skill to declare the tool in the agent prompt, pick MastraD1MemoryStorage for persistence, and avoid the SearXNG engine-name pitfalls. ## Quick Start Ask the agent to wire up a cited search-and-answer handler using createSearchHandlers with your own searchWeb implementation and explain which storage backend fits your deployment.
