LLM & Agents

Implements Anthropic's Parallelization pattern with two modes - Sectioning (independent subtasks in parallel) and Voting (multiple approaches evaluated). Use when tasks decompose into independent units or when comparing implementation strategies. Delivers 2-10x speedup or consensus-based decisions.

Write and debug Bun tests with proper mocking, coverage, and isolation. Use when writing tests, debugging test failures, setting up test infrastructure, mocking fetch/modules, or improving test coverage.

Create, evaluate, and improve Agent skills to production quality (100/100). Use when the user wants to create a new skill, review an existing skill, score a skill against best practices, or improve a skill's quality. Also use when the user mentions skill development, skill templates, or skill optimization.

Guides through adding new features, MCP tools, analyzers, and extending the patent creator system.

Use PROACTIVELY when working on multiple branches simultaneously or creating parallel Claude Code sessions. Automates git worktree creation with prerequisite checking, development environment initialization, and safe cleanup. Supports single worktree, batch creation, and worktree removal. Not for non-git projects or branch management without worktrees.

Craft optimized image generation prompts for tacosdedatos illustrations. Use this skill when you need to create a prompt for Gemini image generation, select an appropriate illustration mode, or help someone write a prompt for the tacosdedatos visual style. Outputs ready-to-use prompts with negative prompts included.

一組資源,幫助我使用公司喜歡的格式撰寫各種內部溝通文件。當被要求撰寫某種內部溝通文件(狀態報告、領導層更新、3P 更新、公司通訊、常見問題、事件報告、專案更新等)時,Claude 應該使用此技能。

Use PROACTIVELY when validating Claude Code skills before sharing or public release. Automated testing framework using multiple isolation environments (git worktree, Docker containers, VMs) to catch environment-specific bugs, hidden dependencies, and cleanup issues. Includes production-ready test templates and risk-based mode auto-detection. Not for functional testing of skill logic or non-skill code.

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.

Generates ready-to-use Phase 0 (Project Setup) prompts for Claude Code. Use when initializing new TypeScript/Node.js projects, bootstrapping project structure, or creating the first phase of an implementation plan. Triggers include "create phase 0", "project setup prompt", "initialize typescript project", "bootstrap project", "init project".

Optimize the research agent for rate limit handling, API call efficiency, web search integration fixes, and improved streaming UX with granular progress updates and source attribution.

Guide for implementing sortable and droppable components using dnd-kit library. Use this skill when building React applications that require drag-and-drop functionality with both container reordering (useSortable) and item dropping (useDroppable) capabilities, such as Kanban boards, file management systems, or playlist editors.

Review code changes (diffs, PRs, patches) and provide structured, actionable feedback on correctness, maintainability, and test coverage. Use when the user asks for a code review, requests feedback on a patch/PR, or wants an assessment of changes.

Find and answer unanswered GitHub discussions. Activate when user wants to contribute to open source discussions, help answer GitHub questions, or find discussions to answer. Delegates all work to subagents.

Flow Nexus user management and system utilities. Use for profile management, storage operations, real-time subscriptions, and platform administration.

Launch Claude Code tasks in new iTerm2 windows for parallel work.USE WHEN user says 'parallel task', 'spawn claude', 'new window task','*parallel', 'launch task', or needs to run a Claude task independently.Supports optional 'happy' mode for mobile continuation via Happy Coder.

Use when creating new skills, editing existing skills, or verifying skills work before deployment - applies TDD to process documentation by testing with subagents before writing, iterating until bulletproof against rationalization

Trace agent decision-making, tool selection, and reasoning chains

Use for token-expensive operations requiring multi-file analysis - codebase exploration, broad searches, architecture understanding, tracing flows, finding implementations across files. Uses MCP cursor-agent server (company pays) with clean async interface. Do NOT use for single-file analysis, explaining code already in immediate context, or pure reasoning tasks.

Automatically log meaningful insights, discoveries, and decisions during coding sessions. Capture non-trivial learnings about edge cases, design decisions, patterns, and implementation details. Activate when building complex code where institutional knowledge should be preserved. Log to .claude/insights/ and help transfer key learnings to CLAUDE.md.