LLM & Agents

LLM fine-tuning with LoRA, QLoRA, and instruction tuning for domain adaptation.

Multi-agent development workflow system for structured task execution. Use when working with tasks in .tasks/ directory, when user invokes /task, /explore, /research, /spec, /build, /review, or /refactor commands, or when Claude needs to understand the task orchestration system. Provides workflow state management, sub-agent coordination, and session persistence.

Integrate external REST and GraphQL APIs with proper authentication (Bearer, Basic, OAuth), error handling, retry logic, and JSON schema validation. Use when making API calls, database queries, or integrating external services like Stripe, Twilio, AWS. Achieves 10-30x cost savings through direct execution vs LLM-based calls. Triggers on "API call", "REST API", "GraphQL", "external service", "API integration", "HTTP request".

強制執行「先讀後寫」規範，確保 Claude 在修改任何檔案前完整閱讀相關代碼。當需要修改檔案時自動使用。

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Claude 4.5 Opus 기반 프로젝트 전체 검수 스킬

Configure OADP for AWS with S3 storage, EBS snapshots, IAM policies, and support for standard clusters and ROSA (Red Hat OpenShift Service on AWS).

Evaluate and improve Claude skills for clarity, completeness, testability, and architectural coherence.

Security patterns for RAG and CAG systems with multi-tenant isolation. Use when building retrieval-augmented or cache-augmented generation systems that require tenant isolation, access control, and secure data handling.

Control Autodesk Fusion 360 CAD software via the Claude Bridge add-in. Create sketches, draw shapes, extrude 3D objects, and build complex models interactively.

Expert on Anthropic Claude API, models, prompt engineering, function calling, vision, and best practices. Triggers on anthropic, claude, api, prompt, function calling, vision, messages api, embeddings

Build AI-native applications with AI SDK v5. Use when integrating LLMs (OpenAI, Anthropic, Gemini) for text generation, streaming, structured outputs, embeddings, image generation, or audio processing. Covers setup, core workflows, provider selection, and advanced features like reasoning, caching, and tool usage.

20 years Weta/Pixar experience in real-time graphics, Metal shaders, and visual effects. Expert in MSL shaders, PBR rendering, tile-based deferred rendering (TBDR), and GPU debugging. Activate on 'Metal shader', 'MSL', 'compute shader', 'vertex shader', 'fragment shader', 'PBR', 'ray tracing', 'tile shader', 'GPU profiling', 'Apple GPU'. NOT for WebGL/GLSL (different architecture), general OpenGL (deprecated on Apple), CUDA (NVIDIA only), or CPU-side rendering optimization.

Transform prompts into structured TCRO format with phase-specific clarification. Automatically invoked by /ai-eng/research, /ai-eng/plan, /ai-eng/work, and /ai-eng/specify commands. Use when refining vague prompts, structuring requirements, or enhancing user input quality before execution.

Implementar clientes de infraestrutura em infrastructure/. Use quando criar client externo, configuração, integração LLM, tracing, ou conexão com serviços fora do synth-lab.

DevOps expertise for deployment, CI/CD, infrastructure, and automation

Detect and fix violations of project instructions defined in .claude/rules/. Use when checking code compliance, reviewing changes, or when the user asks about instruction violations.

This skill should be used when Claude Code needs to perform basic arithmetic calculations. It provides a Python script that safely evaluates mathematical expressions including addition, subtraction, multiplication, division, exponentiation, and square roots.

CRITICAL - Guide for using Claudish CLI ONLY through sub-agents to run Claude Code with OpenRouter models (Grok, GPT-5, Gemini, MiniMax). NEVER run Claudish directly in main context unless user explicitly requests it. Use when user mentions external AI models, Claudish, OpenRouter, or alternative models. Includes mandatory sub-agent delegation patterns, agent selection guide, file-based instructions, and strict rules to prevent context window pollution.

指定したプラグインを有効化する。「プラグインを有効化」「〇〇をオンにして」「〇〇を有効にして」「プラグインを有効に」「〇〇を使えるようにして」「プラグインをオンに」「〇〇を有効化して」などで起動。claude plugin enable コマンドを使用して有効化。