Creates Agent Skills for Claude Code following official standards. Use when the user wants to create a Skill, teach Claude new capabilities, add specialized knowledge, or create reusable instructions. Keywords: create skill, new skill, teach Claude, add capability, skill creation.
Discovers all Claude Code agents in the system including built-in, plugin, project, and user-level agents. Use when you need to find which agents are available, understand the agent ecosystem, or prepare agents for Actoris registration.
このスキルは外部のAI エージェントを実行するためのスキルです。 実装や調査の作業を別のAIエージェントに委譲してください。 あなた（メインのClaude）はユーザーとの対話・設計判断・方針決定に専念し、 実際の調査や実装はこのスキルに任せることでトークンとコンテキストを大幅に節約できます。 重要：実装や調査タスクが発生したら、詳細を考える前にまずこのスキルを使ってください。 実装の細かいやり取りをメインの会話で行うのはトークンの無駄になります。 また、あなたが中間に挟まることで、文脈を省略しがちな人間とタスク実行者との間での口論を防ぐこともできます。
Build Retrieval-Augmented Generation (RAG) systems for LLM applications with vector databases and semantic search. Use when implementing knowledge-grounded AI, building document Q&A systems, or integrating LLMs with external knowledge bases.
Implementing human override capabilities for AI decisions with proper tracking, auditing, and feedback loops.
Build FastAPI backend for OpenAI ChatKit with SSE streaming, conversation persistence, and AI agent integration. Handles /chatkit endpoint, ChatKit-compatible SSE format, conversation models, and message storage. Use when implementing chat backend, SSE streaming endpoint, or connecting AI agent to ChatKit frontend.
Build AI agents with the Claude Agent SDK (TypeScript/Python). Covers creating agents, custom tools, hooks, subagents, MCP integration, permissions, sessions, and deployment. Use when building, reviewing, debugging, or deploying SDK-based agents. Invoke PROACTIVELY when user mentions Agent SDK, claude-agent-sdk, ClaudeSDKClient, query(), or building autonomous agents.
Prompt Engineering Instructional Language (PEIL) - generates optimised system prompts for AI agents and LLMs. Use when creating agent system prompts, improving prompt quality, applying research-backed prompting techniques, or structuring prompts with role, context, instructions, and desired output. Keywords: prompt engineering, system prompt, agent prompt, LLM optimization, prompt generation.
Create or update repo skills under .claude/skills. Use when adding new skills, editing SKILL.md, or deciding what should be a skill vs a spec/doc.
Tune WaveCap Whisper model settings. Use when the user wants to change model size, adjust decoding parameters, configure prompts, or optimize transcription accuracy vs speed.
Nano Banana（Gemini のネイティブ画像生成）向けに、用途（ロゴ/EC物撮り/サムネ/ステッカー/文字入り/ミニマル/差分編集/参照画像固定など）をヒアリングし、 "タグ盛り"ではなく撮影・制作ブリーフ形式で一貫性の高い高品質プロンプトを作るSkill。 ユーザーが「nanobananaで画像生成のプロンプトを作って」「Gemini画像生成の高品質プロンプト」「編集で詰めたい」「参照画像で同一性を保ちたい」等を求めたときに発動する。
Analyzes history/ directories to extract patterns, writes learned rules to .claude/rules/learned/ (auto-loaded), updates CLAUDE.local.md with personal preferences.
Complete SDK for controlling Reachy Mini robot - head movement, antennas, camera, audio, motion recording/playback. Covers architecture (daemon/client), deployment modes (USB, wireless, simulation, on-Pi), and app distribution. Also includes advanced application patterns: MovementManager, layered motion, audio-reactive movement, face tracking, LLM tool systems, and OpenAI realtime integration. Use when: (1) Writing code to control Reachy Mini, (2) Moving the robot head or antennas, (3) Accessing camera/video, (4) Playing/recording audio, (5) Recording or playing back motions, (6) Looking at points in image or world space, (7) Understanding robot capabilities, (8) Connecting to real or simulated robot, (9) Building conversational AI apps, (10) Integrating with LLMs/OpenAI, (11) Deploying apps to robot, (12) Any robotics task with Reachy Mini.
Implement cost-optimized LLM routing with NO OpenAI. Use tiered model selection (DeepSeek, Haiku, Sonnet) to achieve 70-90% cost savings. Triggers on "LLM costs", "model selection", "cost optimization", "which model", "DeepSeek", "Claude pricing", "reduce AI costs".
Use when web.run fails to parse a webpage and you need raw Markdown via Crawl4AI.
Generate effective prompts based on task descriptions. Use when you need to create structured prompts for AI models.
Recover from crashed, failed, or interrupted Claude Code sessions. Use this skill when: session crashed during multi-agent parallel execution, need to determine what work was completed vs incomplete, want to generate resumption commands for interrupted tasks, recovering from context window exhaustion, or handling session handoffs. Analyzes agent logs, verifies on-disk state, and creates resumption plans with ready-to-execute Task() commands.
Guide for building AI agents with Google ADK (Agent Development Kit). Use when creating multi-agent pipelines, implementing conditional agent branching, designing agent tools with FunctionTool, or debugging agent data flow issues. Covers SequentialAgent, LoopAgent, ParallelAgent patterns, session.state management, output_key chaining, and transfer_to_agent for control flow. Essential for understanding non-obvious ADK behaviors like why SequentialAgent runs ALL agents even after rejection.
Comprehensive guide to deploying and serving LLM models including optimization, batching, caching, and production infrastructure
Debug failed/issue tasks from Beam.ai using Langfuse traces. Load when user says "debug issue tasks", "check failed tasks", "why did task fail", "task errors", "debug agent", or needs to investigate task failures.
