Browse reusable Agent Skills, each with a clear purpose and practical guidance.
聚合 .agents/skill-journal/ 中算子开发反馈，生成 skill 改进建议表，开发者命令行勾选后应用到对应 SKILL.md。覆盖所有 .agents/skills/ 下的 skill，不限于 op-design / op-generate。触发：skill review、skill 评审、检查 skill 反馈、应用 skill 改动、tilelang skill 改进、skill 反馈。
Designs AI-powered API features, LLM tool/function definitions, MCP server tool schemas, natural language to API conversion, and agentic API workflows. Use whenever the user asks about "AI calling my API", "function calling schema", "tool definition for LLM", "MCP tools", "natural language API", "AI agent", "let Claude use my API", "OpenAI function calling", "Anthropic tool use", "API agent workflow", or "convert user intent to API calls". Triggers on: "tool schema", "function spec", "agentic API", "LLM plugin", "AI integration", "RAG with my API", or "chatbot that calls my API".
Architecture and safe-extension guide for flowfile_core's /ai/* subsystem — the assist/copilot/planner surface map, the FEATURE_FLAG_AI router gate, the litellm lazy-import contract and its enforcing tests, BYOK provider resolution, per-process rate limiting, the prompt-log debugging runbook, the local llama.cpp model, and two maintainer-validated doctrines (prompt edits must be additive not subtractive; wrong-but-recoverable LLM payload shapes get normalized at the executor seam, not the schema or the prompt) — use when adding or debugging an AI route/agent/provider, when an agent tool call is being rejected or looping, when asked to "make the prompt more direct" or tighten AI/system prompts, when the LLM keeps emitting a wrong JSON shape for a tool call, or when investigating what a Flowfile LLM call actually saw/said.
AI Pair Collaboration Skill. Coordinate multiple AI models to work together: one creates (Author/Developer), two others review (Codex + Gemini). Works for code, articles, video scripts, and any creative task. Trigger: /ai-pair, ai pair, dev-team, content-team, team-stop
Speak as a specified character from their current knowledge, voice, and emotional state. Use for skill-only workflows that need in-character conversation, voice discovery, or relationship pressure tests.
Reference skill for building production-ready crw integrations. Covers verb selection, call surfaces (CLI/MCP/REST), post-filtering strategies, context-window hygiene, Hybrid RAG patterns, common pitfalls, and crw-specific operational considerations (search backend limits, renderer pool, proxy rotation). Load this when writing application code that embeds crw, designing a multi-step agent workflow, or debugging an integration that isn't behaving as expected.
The concierge proxy — turn a vague user script ("I need the agent to help me do X") into a context contract, route it to the right skills/workflows, add backpressure (tests/evals/reviews/approvals), execute, and report. Use when a request is multi-step, durable, or human-in-the-loop and you'd otherwise hand-roll the orchestration; skip it for a single prompt → single answer.
Design a single high-quality prompt — the innermost layer an agent reads. Use when a Smithers <Task>'s prompt (its .mdx body or inline string) is vague, ambiguous, or underperforming and you want to tighten the instruction, role, constraints, examples, and output contract before reaching for harness or workflow changes.
Design the Zod output schema of a Smithers <Task> as the contract between steps. Use when a step's output feeds a later step (or a branch/loop condition) and must be reliable — design the schema first, keep it minimal, and prefer typed fields over prose so downstream rendering can depend on it.
Drive Smithers, a durable control plane for long-running coding agents. Use when the user wants multi-step, long-running, crash-safe, or human-in-the-loop agent work: "orchestrate agents", "run a workflow", "implement this and review it", "keep iterating until tests pass", "plan then build", or anything that needs retries, approvals, replay, or evals across multiple AI steps. YOU (the agent) run Smithers on the user's behalf; it is not a GUI the human clicks. You are an ORCHESTRATOR: run long-running, multi-step, or background work *through* Smithers, not through your own ad-hoc subagents; spend your time observing the run and reporting.