LLM 與 Agent

Strategies for continued pretraining and domain adaptation in Unsloth (triggers: continued pretraining, CPT, domain adaptation, lm_head, embed_tokens, rsLoRA, embedding_learning_rate).

Expert guidance for DeepAgents framework - simplified agent creation with tool integration for LangChain/LangGraph workflows.

Expert RPG map creation and generation for tabletop and online games. Specializes in creating visual descriptions and optimized prompts for z-image turbo model in ComfyUI. Covers all map types (world/continent, regional/local, battle/encounter, dungeon/interior) across multiple themes and settings (fantasy, sci-fi, cyberpunk, post-apocalyptic, horror, etc.). Use when Claude needs to: (1) Create RPG maps of any type, (2) Generate image prompts for map visualization in ComfyUI, (3) Design environments, biomes, settlements, or tactical battlefields, (4) Provide expert knowledge of cartography, terrain, and world-building for games, (5) Convert map descriptions into optimized prompts for image generation models, or (6) Help with any map creation or visualization for tabletop RPGs, video games, or online play.

A Claude Code skill for detecting flaky test patterns that cause intermittent CI failures.

Google File Search API templates, configuration patterns, and usage examples for managed RAG with Gemini. Use when building File Search integrations, implementing RAG with Google AI, chunking documents, configuring grounding citations, or when user mentions Google File Search, Gemini RAG, document indexing, or semantic search.

Expert guide for handling file uploads, image optimization, cloud storage (Supabase, S3, Cloudinary), and file management. Use when implementing file upload features or managing user assets.

This skill should be used when reviewing, auditing, or improving existing Claude Code skills to ensure they follow Anthropic best practices, have proper structure, use current domain-specific patterns, and include all necessary resources. It analyzes skill quality, identifies gaps, suggests improvements, and can automatically enhance skills with updated best practices. Trigger terms include review skill, audit skill, improve skill, enhance skill, update skill, check skill quality, skill best practices, fix skill, optimize skill, validate skill structure.

This skill should be used when translating a spec into code using parallel subagents. Use when the orchestrator invokes the implement phase or when executing a well-defined implementation plan.

Build type-safe LLM applications with DSPy.rb - Ruby's programmatic prompt framework with signatures, modules, agents, and optimization

Orchestrator for WebView UI mockup workflow - delegates design iteration to ui-design-agent and implementation scaffolding to ui-finalization-agent. Use when user mentions UI design, mockup, WebView interface, or requests 'design UI for [plugin]'.

cc-plugins 프로젝트 개발 가이드라인, 레퍼런스 패턴, 프로젝트 구조 조회. Use when Claude needs to (1) Look up development principles (P1-P4, DRY, KISS, YAGNI, SOLID), (2) Find reference patterns for Hooks/Agents/Commands creation, (3) Understand project structure and file locations, (4) Learn how to create Skills/Hooks/Agents/Commands, or (5) Delegate to claude-code-guide for official Claude Code documentation

Create, manage, and cleanup git worktrees with Claude Code agents across all projects. USE THIS SKILL when user says "create worktree", "spin up worktrees", "new worktree for X", "worktree status", "cleanup worktrees", or wants parallel development branches. Handles worktree creation, dependency installation, validation, agent launching in Ghostty, and global registry management.

MUST INVOKE this skill when creating executable plans in .prompts/planning/ with Brief → Roadmap → Phases structure. Expert patterns for hierarchical project plans.

Automatically applies when building RAG (Retrieval Augmented Generation) systems. Ensures proper chunking strategies, vector database patterns, embedding management, reranking, and retrieval optimization.

Guide users through creating Agent Skills for Claude Code. Use when the user wants to create, write, author, or design a new Skill, or needs help with SKILL.md files, frontmatter, or skill structure.

Microsoft Pragmatic Rust Library Guidelines. Use when designing library crates, public APIs, managing dependencies, or creating reusable components.

Use RepoPrompt CLI for token-efficient codebase exploration

Implement and debug semantic/vector search with model2vec-rs and sqlite-vec. Use for embedding generation, vector queries, hybrid search, or RRF fusion.

Process aviation fuel surcharge bills from Excel files. Supports both automatic detection and Claude-assisted modes. Detects headers, matches columns, parses dates/routes, and fetches contract numbers via API. Use when working with aviation fuel bills, Excel file processing, or when user mentions fuel surcharges,航空燃油账单, or 燃油差价费.

Analyze codebase health - large files, test coverage gaps, duplicate code, dead/legacy code, and documentation issues. Use when asked to scan, audit, or assess code quality, find refactoring targets, or identify technical debt.