Use when starting any conversation - establishes mandatory workflows for finding and using skills, including using Skill tool before announcing usage, following brainstorming before coding, and creating TodoWrite todos for checklists
Explain what FAF is, why it matters, and how it works when user asks about FAF, project context, AI-readiness, The Reading Order, or persistent context. Use when user says "what is FAF", "explain project.faf", "why do I need this", "how does AI context work", or shows confusion about persistent context. Teaches foundational concepts before recommending specific actions.
Analyze emotional tone, sentiment polarity, and psychological impact of text. Execute Python script for detailed sentiment analysis. Use when analyzing mood, attitude, or emotional content.
Analyzes user questions and automatically dispatches optimal agents/skills/plugins
Generates anime/manga prompts for Midjourney Niji mode with 30+ artist styles, SREF code library
Multi-language greetings in 6 languages. Use for non-English greetings or multiple people.
將使用者的自然語言需求，轉換成「Codex 可直接執行」的工程指令 prompt。 輸出內容包含：背景、明確目標、可驗收條件、操作步驟、不可違反的限制、與交付格式。 適用於任何語言、任何 repo、任何專案階段。
Designing effective prompts - system/user prompts, few-shot learning, chain-of-thought, defensive prompting, injection defense. Use when crafting prompts, improving outputs, or securing AI applications.
Curates context, optimizes prompts with XML, and manages extended thinking for Anthropic Claude models. Use when building Claude-based agents, designing system prompts, or handling long-context tasks.
This skill should be used when the user asks to "configure a subagent", "create a custom agent", "set up agent tools", "configure permissionMode", "add agent skills", or mentions subagent fields like tools, model, or permissionMode. Provides comprehensive guidance on subagent configuration options and best practices.
Generate a skeleton template for a new Claude Code Skill.
Connect to the Pieces MCP server (SSE) and reliably query or write to Pieces Long‑Term Memory (LTM) using query/write tool patterns (e.g., ask_pieces_ltm + create_pieces_memory), with practical troubleshooting and request-shaping examples.
Respond with different character personalities (pirate, butler, professor) when the user requests character-style responses. Use when the user says phrases like "talk like a pirate", "respond as a butler", "explain like a professor", or similar requests in Japanese or English.
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
The Orchestrator Agent - Master coordinator for the Nebuchadnezzar v4.0 Pipeline OS. Manages the Four Pillars: Expert CRUD, Observability, ADHD Execution Loop, and Learning Propagation. This agent ensures all domain experts stay aligned, improve continuously, and execute the right tasks. Use when: (1) Starting a new session - runs HUD check and recommends focus (2) "orchestrate" or "coordinate" - multi-expert task routing (3) "which expert" or "load expert" - expert selection guidance (4) "system health" or "pipeline status" - full observability report (5) "propagate learning" or "update experts" - cross-expert knowledge sharing (6) Any red stage in HUD - automatically routes to correct expert Triggers on: "orchestrate", "coordinate", "system health", "which expert", "load expert", "propagate", "adhd loop", "four pillars", "pipeline os"
Search claude-mem's persistent cross-session memory database. Use when user asks "did we already solve this?", "how did we do X last time?", or needs work from previous sessions.
Use when multiple independent tasks can run simultaneously. Enables efficient parallel work execution with specialized agents.
Use when ending a session and want to spawn a continuation session in agent-deck with inline context
Use when starting any conversation - establishes how to find and use skills, requiring Skill tool invocation before ANY response including clarifying questions
Expert guidance for LangGraph Python library. Build stateful, multi-actor applications with LLMs using nodes, edges, and state management. Use when working with LangGraph, building agent workflows, state machines, or complex multi-step LLM applications. Requires langgraph, langchain-core packages.
LLM-as-Judge techniques including direct scoring, pairwise comparison, rubric generation, and bias mitigation.
This skill should be used when the user asks to "create a prompt", "optimize a prompt", "improve this prompt", "engineer a prompt", "prompt engineering best practices", "make this prompt better", "recommend a model", "which model should I use", "best model for", "GPT vs Claude", "Opus vs Sonnet", "Haiku vs Sonnet", "analyze prompt quality", "fix my prompt", "prompt for Claude", "prompt for GPT", or needs help with prompt engineering techniques, model selection, or prompt optimization for any LLM (Claude Opus/Sonnet/Haiku 4.5, GPT 5.1/Codex, Gemini Pro 3.0).
A minimal example SKILL for OpenCode. When executed it simply returns the string "Hello, OpenCode!". This can be used to test SKILL loading and execution.
