Use when the user wants to design MCP tool candidates, function/tool calling schemas, or agent tool risk gates from project API surfaces. Triggers on phrases like "/knowject mcp tools", "design MCP tools from this API", "把这些 endpoint 设计成 MCP tools", "为 agent 工具加确认 gate", "API-to-MCP tool design", or "which endpoints are safe for tool calling". Day-1 output is design artifacts only under knowject/tools/: README.md, mcp-tools.plan.md, mcp-tools.schema.json, and tool-risk-report.md. It does not implement an MCP server or runtime tool handlers.
Use when the user wants to create or review lightweight RAG, retrieval, and citation-quality evaluation cases from project knowledge artifacts. Triggers on phrases like "/knowject rag eval", "evaluate RAG citations", "生成 RAG 评测用例", "检查引用是否有来源", "把这些知识库文档做成 eval cases", or "评估事实和 roadmap 是否混淆". Day-1 output is file-based only under knowject/evals/: README.md, rag-eval-cases.yaml, and rag-eval-report.md. Every eval case must include source evidence, avoid secrets, and pass a diff-confirm gate before writing.
Use when the user wants to generate TypeScript types from an OpenAPI document and wire them into the typed client produced by knowject-read-api. Triggers on phrases like "/knowject types", "把 openapi 转成 types", "给 read-api client 补类型", "生成 typed client 的类型", "rewrite the unknown placeholders", "wire types into the generated client". Day-1 input is OpenAPI 3 only — non-OpenAPI inputs (Express routes, Mongoose, Zod, Pydantic, Spring DTOs, Nest decorators, markdown specs, etc.) are refused with framework-specific recommendations for exposing OpenAPI. Two output modes: (1) primary — emit a sibling <module>.types.ts and rewrite request<unknown>(...) in the read-api client to request<X>(...); (2) fallback — emit types-only when no read-api client file is found. Both modes use the diff-first / confirm-before-write safety gate established in Phase 2.
Use when the user wants to set up Knowject Skills in a project, initialize knowject/context.yaml, or onboard a project to cross-role collaboration Skills. Triggers on phrases like "/knowject init", "初始化 knowject", "set up knowject skills", "在这个项目里启用 knowject", "knowject 配置", "create context.yaml", "搭一下 knowject", or natural intent like "I want to use Knowject Skills here". Detect when knowject/context.yaml is missing and the user expresses a wish to use any other knowject-* Skill — then proactively offer to run this first.
Use when the user wants to capture durable, source-cited project memory from repo artifacts, handoff notes, diffs, PR notes, or conversation summaries. Triggers on phrases like "/knowject memory", "沉淀一下这个项目的记忆", "把这次交接总结成 knowject memory", "capture project memory", "update knowject memory from these docs", or "把这次 PR / diff / 对话沉淀成项目记忆". Day-1 output is file-based only: knowject/memory/README.md and knowject/memory/project-memory.yaml. Every memory item must include source evidence, avoid secrets, and pass a diff-confirm gate before writing.
Use when the user asks to discover backend API endpoints or generate a typed HTTP client in a project that has knowject/context.yaml. Triggers on phrases like "/knowject api", "找一下用户列表 endpoint", "生成 typed client", "has there been a POST /orders endpoint", "scan the api", "wire up the client for the users module", "把 user 接口的 client 帮我写一下". Two modes: discovery (answer questions about the API surface) and generation (emit a typed-client TypeScript file under api.client.output_dir using api.client.wrapper). Reads api.sources[*] from knowject/context.yaml — Phase 2 supports format=express and format=openapi.
Use when the user wants to turn a written product requirement into a single-page HTML mock anchored to the project's brand. Triggers on phrases like "/knowject mock", "把这段 PRD 出一个原型", "做个高保真 mock", "generate a mock from this spec", "用我们的品牌色出个原型", "我描述一下这个页面你出 HTML". Reads brand tokens (primary_color, font_family, voice, logo_path) and locale from knowject/context.yaml — output looks like the user's product, not a generic AI Tailwind page. Writes one self-contained .html file (no build required); never overwrites silently.
Use when the user wants to turn a UI design source (PNG screenshot, PDF page, Figma export, or sketch) into component skeleton code aligned to the project's UI library and file conventions. Triggers on phrases like "/knowject design", "把这张设计稿拆成组件", "convert this mockup to React components", "generate component skeletons from this design", "我贴一张图你出 antd 组件骨架". Reads stack.frontend (framework, ui, styling, language) and design.output from knowject/context.yaml. Two-phase: outputs a decomposition plan first for user confirmation, only then writes skeleton files. Never overwrites existing components silently. Phase 2 supports antd / mui / shadcn / chakra / naive / mantine UI libraries; other values fall back to plain React + Tailwind.