vector-search-patterns
Implementing semantic search and similarity search using vector embeddings and vector databases.
Implementing semantic search and similarity search using vector embeddings and vector databases.
Automatically applies when designing multi-agent systems. Ensures proper tool schema design with Pydantic, agent state management, error handling for tool execution, and orchestration patterns.
Persistent memory system for Claude Code. Captures discoveries, tech debt, and unfinished work that persists across sessions. Use for cross-session continuity and sharing findings with team. Triggers: "remember this", "capture finding", "what did I find", "session context", "load context", "save session", "findings", "discoveries", "tech debt to track", "don't forget", "note for later", "carry forward", "handoff notes", "open findings", "ready to work on".
TOON Format Specialist - Token-efficient data encoding for LLM communication optimized per TOON Spec v2.0
AI导师框架,通过10+顶级提示工程方法论(C.R.E.A.T.E.、Perfect Prompt Framework、修辞学方法、结构化方法、元提示词理论、迭代优化、个性化定制、逆向思维、系统思维、多模态应用)引导用户创建高效元提示词。当用户需要:(1) 学习提示工程最佳实践,(2) 创建特定任务的元提示词,(3) 优化现有提示词效果,(4) 掌握提示设计思维模式,(5) 建立个性化提示词库,(6) 应用框架解决实际问题,或 (7) 发展AI协作技能时触发。通过'创建元提示词'、'优化提示'、'提示工程指导'、'AI导师'等关键词激活。
Use the Anthropic Claude SDK for programmatic API access. Enables spawning sub-agents, parallel processing, batch operations, and custom Claude interactions via Python. Use when tasks need multiple Claude instances, background processing, or direct API control.
AI 創意教練腦力激盪。當用戶說「brainstorming」、「腦力激盪」、「創意思考」、「想點子」、「發想」、「brainstorm」、「一起來想」、「幫我想」、「集思廣益」、「ideation」、「創意發想」時啟動。提供十大類別創意技術庫,由 AI 智能推薦最適合的組合,以教練式對話引導創意探索。適用於:創新發想、問題解決、策略規劃、團隊凝聚、打破框架、個人洞察。
Token optimization and context window management strategies. Load for long sessions, complex tasks, or when approaching context limits.
Work with Leavn search system - UnifiedSearchService, FTS5 database, 50ms debounce, content filtering, search history
Expert guidance for Python AI development fundamentals, essential libraries, data structures, async programming, and best practices for AI/ML projects
Build Retrieval-Augmented Generation systems with vector databases
Generate stunning photorealistic images using advanced AI models. Covers portrait photography, landscapes, product shots, architecture, and cinematic scenes. Includes prompting techniques for lighting, composition, and camera settings. USE WHEN user says 'photorealistic', 'photo realistic', 'realistic image', 'stunning photo', 'cinematic', 'portrait photo', 'product photography', 'landscape photo', 'hyper realistic', 'professional photo', or needs high-quality realistic image generation.
Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions
Helper skill to create new agent skills following the standard structure. Use this when you want to define a new capability or workflow for the agent.
Persistent iteration wrapper for cognitive reasoning patterns using ralph-loop's Stop hook mechanism. Use when high confidence (>90%) is required, complex multi-pattern orchestration needs iterative refinement, self-correcting analysis is needed, or long-running tasks require checkpointed persistence. Wraps IR-v2 patterns in completion promise-gated loops.
Generates high-quality, predictable, and efficient .agent/skills/ directories based on user requirements. Use when the user asks to create a new skill.
Creates a new skill in the Powerhouse Lab skills library with proper git workflow. Use when the user wants to create a new skill, add a skill, build a skill, or says "new skill called X".
Creates new Agent Skills with proper YAML frontmatter, progressive disclosure architecture, and best practices. Use when the user asks to create a new skill, generate a skill template, or build custom capabilities.
Nemotron-Personas-Japan (1M件) から最適ペルソナを選定し、Instagram実データで検証・統合する全自動ペルソナ分析Skill。Instagram分析、ペルソナ作成依頼時に自動起動。