LLM & Agents

新機能調査、提案、設計方針確認時に使用。

SQLite database for Israeli Tech Nano-Influencers with direct sqlite3 command-line access. Execute any SQL query or inspect schema. Simple and agent-friendly.

How to spawn subagents correctly. Use this reference when you need to delegate work to a specialized agent.

Comprehensive LLM model evaluation and ranking system. Use when users ask to compare language models, find the best model for a specific task, understand model capabilities, get pricing information, or need help selecting between GPT-4, Claude, Gemini, Llama, or other LLMs. Provides benchmark-based rankings, cost analysis, and use-case-specific recommendations across reasoning, code generation, long context, multimodal, and other capabilities.

Precise token counter for GPT, Claude, and Gemini models (source of truth from software kernel)

Optimize documentation for conciseness and clarity by strengthening vague instructions and removing redundancy. Use when asked to optimize, condense, or clarify documentation files, Claude Skill, Claude Command, etc.

Expert in managing Kubernetes clusters using kubectl-ai and kagent. Use this for generating Helm charts, troubleshooting pods, and automating cluster operations.

Claude 스킬을 생성하고 작성하는 것을 도와줍니다. 새로운 스킬 만들기, 스킬 작성 가이드, 베스트 프랙티스 적용 등의 키워드에 반응합니다.

AI-powered generative UI with Thesys - create React components from natural language.

Activates when reviewing code to identify quality issues, security vulnerabilities, and suggest improvements

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.

Prompt engineering technique to overcome mode collapse in LLM responses by generating multiple answers with probabilities. Use when you want to see alternative responses beyond the top-ranked answer, explore the full distribution of possible responses, discover unexpected solutions, or understand the probability range of different answers. Particularly useful for creative tasks, complex decision-making, brainstorming, and situations where conventional top-ranked answers may be limiting.

Patterns for wrapping any agent with RAG context from Qdrant. Use to add persistent memory to imported or external agents.

Expert assistant for managing Finnish/English content translations in the KR92 Bible Voice project. Use when syncing topic translations, managing translation cache, validating translation completeness, or importing/exporting CSV translations.

search the web for information when user explicitly requests web searches, by identifying the nature of the request and delegating to appropriate subagents (@code-researcher for documentation/code examples, @web-researcher for general information)

Manage Gemini RAG stores with Code-RAG support

Generate pytest test templates for UTXOracle modules following TDD patterns. Automatically creates RED phase tests with async fixtures, coverage markers, and integration test stubs.

This skill should be used when choosing between flags, settings, or files for data storage, implementing document flags, registering module settings, handling file uploads, or migrating data between storage types. Covers namespacing, scope types, and performance optimization.

Spawns 4 parallel review agents (Accessibility, React Router, React Core, Anders Hejlsberg) to conduct comprehensive multi-perspective code reviews

Clean Architecture and Cosmic Python guidance for well-tested, layered Python systems. Use for designing Python projects with layered architecture (models, adapters, services, entrypoints), enforcing Clean Code and SOLID principles, testing strategies (unit tests, BDD, Gherkin), CI/CD setup (pytest, tox, importlinter), and architectural decision-making (ADRs). Applicable to systems requiring strict boundary enforcement, clean separation of concerns, and comprehensive test coverage.