agent-authoring
Guide for authoring specialized AI agents. Use when creating, updating, or improving agents, choosing models, defining focus areas, configuring tools, or learning agent best practices.
Guide for authoring specialized AI agents. Use when creating, updating, or improving agents, choosing models, defining focus areas, configuring tools, or learning agent best practices.
撰写提示词、优化提示词、改写 Prompt、Prompt 优化。将模糊的想法转化为精确有效的 AI 提示词,支持 ChatGPT、Claude、Gemini、DeepSeek、豆包、智谱等多平台。
A universal self-improving agent that learns from ALL skill experiences. Uses multi-memory architecture (semantic + episodic + working) to continuously evolve the codebase. Auto-triggers on skill completion/error with hooks-based self-correction.
Convert MCP servers to typed TypeScript APIs for efficient code execution. Reduces token usage by 98%+ by transforming tool calls into programmatic access. Use when building agents that need to interact with multiple MCP servers efficiently, when context window is a concern, or when native control flow (loops, conditionals) would simplify multi-step workflows.
AG-UI (Agent-User Interaction) protocol reference for building AI agent frontends. Use when implementing AG-UI events (RUN_STARTED, TEXT_MESSAGE_*, TOOL_CALL_*, STATE_*), building agents that communicate with frontends, implementing streaming responses, state management with snapshots/deltas, tool call lifecycles, or debugging AG-UI event flows.
Analyze LangGraph application architecture, identify bottlenecks, and propose multiple improvement strategies
Creates claude-queue tasks when usage limits are approaching. Use when the user is running low on Claude usage quota, mentions hitting limits, or when significant work remains and we should queue it for unattended execution. Proactively suggests queuing remaining work during long coding sessions.
Create custom Task subagents in ~/.claude/agents/. Use when defining new agents for Task tool delegation.
Multi-turn safety persistence evaluation for crisis scenarios. Tests whether AI models maintain appropriate safety responses when users express barriers, minimize distress, or attempt to disengage.
Skill for interacting with the Lucid Client API (multi-agent runtime). The API is hosted in the lucid-client codebase and provides endpoints for managing agents, invoking entrypoints, handling payments, and more. Activate when: user wants to interact with the Lucid Client API, manage agents, invoke agent entrypoints, or work with the multi-agent runtime system.
Guide for creating new Agent Skills with proper structure, frontmatter, bundled assets, and validation. Includes templates, best practices, and examples for building reusable skill resources.
Guide for adding new agents to the ADK pipeline. Use when creating a new LlmAgent, SequentialAgent, or ParallelAgent, or when extending the pipeline with additional processing stages.
Build and deploy production-ready generative AI agents using Vertex AI, Gemini models, and Google Cloud infrastructure with RAG, function calling, and multi-modal capabilities
Get a specific vector entry by key. Requires authentication. Use for Agentuity cloud platform operations
Complete reference for integrating with 300+ AI models through the OpenRouter TypeScript SDK using the callModel pattern
Long-term user profile memory that persists across Claude Code sessions
Sage Prompt 工程规范,学习 Crush 的模板化方案,包含 system prompt 设计最佳实践
Guide for using LLM utilities in speedy_utils, including memoized OpenAI clients and chat format transformations.