Global hooks for logging Claude Code conversation events to markdown files. Tracks prompts, tool usage, and responses across all sessions. Useful for debugging, auditing, and providing conversation context to Claude.
Session lifecycle management for autonomous coding. Use when starting sessions, resuming work, detecting session type (init vs continue), or managing auto-continuation between sessions.
從 YouTube、Apple Podcast 或 Bilibili 的 URL 獲取逐字稿。 當使用者貼上影片或 podcast 連結，或說「幫我拿逐字稿」「轉成文字」時觸發。 使用 aits 命令獲取內容，回傳給使用者後續使用。
Helps IDE agents navigate and operate the MOVA Skills Lab: read the registry, plan skill runs and manage episodes.
When working with JSON data in LLM prompts (especially large arrays or tabular data), consider the token-efficient TOON (Token-Oriented Object Notation) format which reduces tokens by 30-70% while maintaining lossless JSON representation and structural validation. Use for reading/writing .toon files, converting JSON↔TOON, or optimizing structured data for LLM consumption with guardrails like [N] counts and {field} headers.
Manage and compact context for long sessions. Use when context is filling up, creating handoff summaries, optimizing context usage, or preparing for session continuation.
智能 Prompt 优化顾问技能。基于 Anthropic 官方最佳实践,通过场景识别和技术匹配,将用户的模糊需求转化为高质量 prompt。触发场景:(1) 用户明确请求"优化 prompt"、"帮我写 prompt"、"改进这个指令";(2) 用户的需求描述模糊、不清晰、缺少关键信息;(3) 用户提到"不知道怎么表达"、"效果不好"等问题。适用于所有需要与 Claude 交互的场景,通过结构化对话挖掘需求,生成符合 Claude 4.5 最佳实践的优化 prompt。
Suno用のBGMプロンプトを生成し、YouTubeにアップロードするためのAgent skill
Orchestrate reliable multi-agent reasoning using MAKER (Maximal Agentic Knowledge Engine for Reasoning). Implements three-pillar architecture for transforming probabilistic LLM outputs into deterministic, verifiable results. Use when tasks require high reliability, parallel consensus voting, or systematic error detection. Triggers include reliability-critical tasks, multi-step reasoning chains, consensus-based verification, parallel agent execution, or explicit MAKER invocation.
[11] CONTINUE. Resume an existing session. Use when continuing previous work to load AGENTS.md, MEMORY.md, and the chosen session log, then confirm readiness.
Debug OpenAI/Gemini image generation failures for Sketch Magic. Use when `/api/convert` returns 500, when models are unsupported, when outputs are empty, or when provider timeouts/quota errors appear.
LangChain implementation patterns with templates, scripts, and examples for RAG pipelines
Interact with TextCleaner's REPL interface to manage pipelines, set input text, and process data through the socket server
Search through historical Claude sessions for code, conversations, and commands. Use when user asks about previous sessions, past conversations, "did I ever", "find in history", or wants to retrieve code/info from earlier chats.
Auto-loaded at session start via SessionStart hook. Teaches skill invocation protocol, tool selection rules (look-at for media, skills for workflows), agent delegation patterns, and enforcement mechanisms. NOT user-triggered - provides foundational skill usage discipline for all sessions.
Use when Codex needs curated documentation context or outlines from the Context7 MCP server.
Creates new specialist subagents based on detected tech stack or observed patterns. Use proactively during /claudenv to create agents for detected technologies, or when usage patterns suggest a new specialist is needed. Similar to meta-skill but for agents.
Use when "CLIP", "Whisper", "Stable Diffusion", "SDXL", "speech-to-text", "text-to-image", "image generation", "transcription", "zero-shot classification", "image-text similarity", "inpainting", "ControlNet"
Creates new AI agent skills following the Agent Skills spec. Trigger: When user asks to create a new skill, add agent instructions, or document patterns for AI.
Create, configure, and document custom Claude Code subagents, including choosing scope, defining YAML frontmatter, selecting tools, model, permission mode, hooks, and usage examples.
