LLM 與 Agent

Use this skill when the user asks to create characters, backstories, motivations, arcs, traits, or personality systems for fiction writing.

Coordinate multiple agents by routing tasks to appropriate specialists. EXCLUSIVE to project-manager agent.

Validates code quality before commits in the MigraineTracker React Native project. Use this skill when preparing to commit code, before pushing changes, or when the user asks to run pre-commit checks. Ensures TypeScript type safety, test coverage (80%+), passing tests, and proper branch strategy compliance.

Use when working with Deno edge functions, LLM integration, or embedding generation. Load for Deno.serve patterns, Zod request validation, OpenRouter LLM calls, and error handling. Covers function structure, CORS, and the call-llm/generate-embedding patterns.

Full-featured AI chatbot with Next.js 15, Auth.js, Drizzle ORM, and multi-model support.

Use when creating or modifying Claude Code hooks - hooks configuration with events, matchers, and JSON responses for November 2025 Hooks v2 specifications

Best practices for crafting effective prompts for agents, skills, and Claude interactions.Use when writing system prompts, agent instructions, or optimizing Claude responses.

RBAC/監査ログ/秘密情報/マルチテナント越境/LLMプロンプト注入の観点でレビューする。

Extract learnings from completed swarm work for persistent memory storage. Analyzes swarm_record_outcome data to identify corrections, decisions, and insights. Stores findings in Memory Lane for future reference. Use when subtasks complete with notable learnings or when you want to capture what worked/didn't work.

A personal welcome message for Ricky from his best friend Steven, introducing this custom Linux learning journey. Use this when Ricky first starts exploring Linux or needs encouragement.

Spawn new terminal windows to run Claude Code, Gemini CLI, or raw CLI commands in parallel. Use when the user wants to fork a task, spawn an agent, run parallel work, delegate a task to a new terminal, or mentions "fork", "spawn", "new terminal", "parallel agent".

Validate plugin integrity and offer safe auto-fixes for common issues - checks agents, skills, hooks, routing config, and plugin structure

This skill should be used when the user asks to "use Gemini Batch API", "process documents at scale", "submit a batch job", "upload files to Gemini", or needs large-scale LLM processing. Includes production gotchas and best practices.

Expert prompt engineering for LLMs and AI agents. Use when users request help writing, reviewing, or improving prompts for chatbots, AI agents, system prompts, instruction sets, or any LLM-based application. Applies research-backed techniques to minimize hallucinations and maximize reliability.

Design, review, and improve agent workflows & agent using SSOT, SRP, Fail Fast principles. Supports Prompt Chaining, Parallelization, Orchestrator-Workers patterns.

Generate images using Google's Nano Banana Pro (Gemini 3 Pro Image) with workflow-based prompting

Use PROACTIVELY when you need to create, update, configure, or validate Claude hooks for various events and integrations

Orquestrador de tarefas que coordena agentes especializados. Ativado automaticamente quando um plano de tarefa é aprovado pelo usuário ou quando o usuário fala "orquestre a tarefa", "orquestrar tarefa", "inicie a orquestração". NÃO ativar no modo plano. Delega para dev-executor-senior-enterprise (implementação), playwright-manual-tester (testes manuais) e qa-architect-senior (testes em escala). NUNCA executa tarefas diretamente - apenas orquestra.

Ensure quality through testing, documentation, and security audits. Use when asked to add tests, improve test coverage, update documentation, write docs, security audit, review safety rules, or verify protected paths are not touched.

Expert on Claude Code settings.json configuration, permissions, sandbox, environment variables, and settings hierarchy. Triggers when user mentions settings.json, permissions, allow rules, deny rules, sandbox, hooks configuration, or settings precedence.