LLM & Agents

Automatically save exploration findings, discoveries, and current state to the exploration journal (docs/exploration/). Use when making important discoveries or completing a phase of work. Ensures future Claude sessions can continue where this session left off.

Advanced prompt optimization and composition system for all prompt types (system prompts, task-specific, creative, technical, agentic). Use when users need to refine, optimize, or transform prompts to achieve better AI outputs. Triggers include requests to improve prompts, make prompts more effective, analyze prompt quality, create prompts from scratch, optimize existing prompts for specific goals, or optimize for Claude 4.x models. Also use when users share prompts that could be improved or ask for prompt engineering guidance.

OpenProse is a programming language for AI sessions. An AI session is a Turing-complete computer; OpenProse structures English into unambiguous control flow. More pattern than framework—it ships as a skill with no dependencies. Activate when: running .prose files, mentioning OpenProse, or orchestrating multi-agent workflows from a script. Use this skill if you ever want to kick off more than one subagent at a time, or orchestrate anything interesting between more than one subagent. Write a .prose file and save it in the .claude-plugin/ directory. Then embody the OpenProse VM, as described in prose.md, and execute it.

Build real-time, reactive backend applications with Convex using TypeScript queries, mutations, and actions with automatic reactivity and optimistic updates. Use when building real-time collaborative applications, implementing reactive data synchronization, writing serverless backend functions, creating queries that auto-update, implementing mutations with transactional guarantees, handling file uploads with Convex storage, implementing authentication with Convex Auth, designing reactive database schemas, or building applications requiring instant data synchronization.

Build modern Next.js 13+ applications using App Router architecture with Server Components, Client Components, and advanced routing patterns. Use when implementing server-first rendering, creating nested layouts, building parallel routes, implementing intercepting routes, using React Server Components, optimizing data fetching with server-side async components, implementing streaming with Suspense, managing client-side interactivity boundaries, or leveraging Next.js 13+ app directory features for performant, SEO-friendly React applications.

Systematically review pull requests, feature implementations, and code changes to ensure quality, maintainability, security, and adherence to best practices. Use when reviewing code before merging, conducting peer reviews, performing self-reviews, auditing code quality, checking for security vulnerabilities, ensuring consistent coding standards, verifying test coverage, assessing performance implications, evaluating architectural decisions, or providing constructive feedback to improve team code quality.

Create well-structured, comprehensive Claude skills with clear activation criteria, detailed descriptions, usage examples, and proper documentation following best practices. Use when documenting new workflows, creating reusable skill patterns, establishing coding standards as skills, building team knowledge base, or expanding the skills library with new capabilities.

Premium app generation that creates WOW-factor experiences. Multi-page apps with smooth animations, zero TypeScript errors, and production-ready quality. Lovable-style experience: one prompt, complete app, instant delight. MUST be used alongside vibe-orchestrator for new projects.

Supabase integration specialist. Handles database schema, authentication, Row Level Security (RLS), real-time subscriptions, and storage. Connects existing UI to real backend. Only called AFTER UI exists with mock data. Triggers: connect database, connect Supabase, add auth, make login, backend integration, real data, authentication, database schema.

Master workflow controller for Lovable-style development. Creates working UI immediately from ANY prompt - no questions asked, no choices given. Routes to specialist sub-agents (ui-builder, dev-builder, design-reviewer, backend-connector, platform-adapter). Triggers: create, build, make, want, new project requests, app ideas, MVP, prototype, or any development request. This skill MUST be read first for any development task.

Design system and anti-patterns for professional UI. Ensures apps don't look "AI generated". Defines color palettes, typography, spacing, animations, and common anti-patterns to avoid. Used by Design Reviewer sub-agent to polish UI. Triggers: design review, polish UI, improve aesthetics, looks like AI made it, design system, style guide, make it beautiful.

Design co-learning experiences using the Three Roles Framework (AI as Teacher/Student/Co-Worker). Use when teaching AI-driven development workflows, spec-first collaboration, or balancing AI assistance with foundational learning. NOT for curriculum without AI integration.

Reference manual for framework methodology throughout session. Provides guidance on cognitive architecture that replaces default AI assistant behaviors with systematic patterns proven through collaborative experience. Consult this manual throughout session when applying framework procedures.

Design reusable prompt templates that encode domain-specific patterns for recurring AI tasks. Use when you've executed similar prompts 2+ times and need to capture the pattern as reusable intelligence. NOT for one-off prompts or generic "ask AI a question" patterns.

Design systematic decision frameworks for selecting appropriate AI tools (Claude Code vs Gemini CLI)based on context requirements, codebase size, reasoning depth needs, and task complexity. Use thisskill when starting projects with unclear tool requirements, optimizing context-constrained workflows,or designing multi-phase strategies that leverage multiple tools' strengths. This skill helps matchtool capabilities to task characteristics, preventing wasted context and ensuring optimal resourceallocation.

Creates assessments with varied question types (MCQ, code-completion, debugging, projects) aligned to learning objectives. Use when educators design quizzes/exams, need questions at appropriate Bloom's cognitive levels, want balanced cognitive distribution (60%+ non-recall), or require rubrics for open-ended questions.

Technical design collaboration through natural dialogue. Adapts to expertise level and problem complexity through understanding, exploration, and validation stages. Use when user asks to brainstorm ideas for new features or systems requiring architectural decisions before implementation begins.

Design and create Claude Skills using progressive disclosure principles. Use when building new skills, planning skill architecture, or writing skill content.

Use when creating temporary files, drafts, experiments, or any content that should not be committed to version control. Ensures proper placement in .claude/.scratch with gitignore configuration.

Use when configuring interactive controls and args for Storybook stories. Helps create dynamic, explorable component demonstrations with proper type constraints.