seanchiuai

AI/ML integrations including multi-provider LLM routing, embeddings, RAG pipeline, autonomous agents, Langfuse prompt management, image generation, and observability. Use when working on LLM provider config, adding new AI models, modifying the RAG/embedding pipeline, building agents, changing prompt templates, or integrating image generation providers. Do NOT use for voice-specific features (use voice skill) or frontend UI (use ui-components skill).

Nuxt 3 server API routes, middleware, agents, background queues, MCP server, voice pipelines, and server utilities. Use when creating new API endpoints, modifying server middleware, working with the agent runtime, adding server utils, or debugging API errors. Covers route conventions (GET/POST/PUT/DELETE), auth middleware, error handling, and Nitro server architecture. Do NOT use for background job processing (use processor skill) or database schema changes (use database skill).

Authentication and authorization using Supabase Auth with PKCE flow. Use when working on login/signup, session management, API key auth, OAuth integration, native app auth (Apple/Google Sign-In), or debugging auth errors. Supports web (cookie), native (Bearer token), API key, and OAuth flows. Do NOT use for general API route creation (use api-server skill) or Stripe billing auth (use integrations skill).

Guide for creating new Claude Code skills for the Minds AI webapp. Use when asked to "create a new skill", "add a skill for X", or when documenting a new domain area of the codebase. Covers skill structure, YAML frontmatter format, instruction writing, and skill organization conventions. Do NOT use for OpenClaw agent skills (different format and location).

PostgreSQL database with Prisma ORM, pgvector for embeddings, and pgbouncer connection pooling. Use when modifying database schema, writing migrations, adding new Prisma models, optimizing queries, working with vector embeddings, or debugging database connection issues. Covers schema conventions, migration workflow, RLS policies, and connection management. Do NOT use for Supabase Auth specifics (use auth skill) or background job queues (use processor skill).

DigitalOcean App Platform deployment architecture and procedures for the Minds AI webapp. Use when asked about "deployment", "how to deploy", "staging vs production", "DO app platform", "Dockerfile", "environment variables", "build process", or "deploy pipeline". Covers staging/production environments, Docker multi-stage build, GitHub Actions workflow, and env var management. Do NOT use to actually trigger deploys (deployments are manual-only by Alexander). Do NOT modify DO env vars directly (use GitHub Environments workflow).

Git branching strategy and commit conventions for the Minds AI webapp. Use when creating branches, making commits, opening PRs, resolving merge conflicts, or following the staging-to-production workflow. Covers branch naming (feature/, fix/, chore/), commit message format, PR creation targeting staging, and the rule that only Alexander merges staging to main. Do NOT use for deployment procedures (use deployment skill) or CI/CD pipeline details.

Third-party service integrations including Stripe billing, Google Calendar, Twilio (SMS/WhatsApp/Voice), Resend email, Firebase push notifications, analytics (PostHog/GA/TikTok), Langfuse observability, content scraping (Apify/Tavily/Serper), and image generation (Replicate). Use when adding a new third-party integration, modifying billing logic, working with webhooks, configuring analytics tracking, or debugging external API calls. Do NOT use for AI/LLM provider integration (use ai-ml skill) or voice pipeline specifics (use voice skill).

Build Convex backends with queries, mutations, actions, HTTP endpoints, and schemas. Comprehensive guide for all Convex patterns and workflows.

Build Next.js applications with App Router, Server and Client Components, data fetching, routing, and TypeScript. Use when implementing Next.js pages, layouts, API routes, or React Server Components.

Install and customize shadcn/ui components with Tailwind CSS theming, variants, and styling. Use when adding UI components or customizing the design system.

Automatically deploys to Vercel production, uses Vercel MCP to fetch build logs, analyzes errors, fixes them, and retries until successful deployment. Use when deploying to production or fixing deployment issues.

Create new domain expert with expertise file and workflows. Use when adding new technology or domain area.

Generate prompt variations from templates. Use when creating new commands, skills, or workflows.

Used to create a new agent. Used when a user wants to create a new agent

Used to create a new skill. Used when a user wants to create a new skill

AI/ML integrations including multi-provider LLM routing, embeddings, RAG pipeline, autonomous agents, Langfuse prompt management, image generation, and observability. Use when working on LLM provider config, adding new AI models, modifying the RAG/embedding pipeline, building agents, changing prompt templates, or integrating image generation providers. Do NOT use for voice-specific features (use voice skill) or frontend UI (use ui-components skill).

Nuxt 3 server API routes, middleware, agents, background queues, MCP server, voice pipelines, and server utilities. Use when creating new API endpoints, modifying server middleware, working with the agent runtime, adding server utils, or debugging API errors. Covers route conventions (GET/POST/PUT/DELETE), auth middleware, error handling, and Nitro server architecture. Do NOT use for background job processing (use processor skill) or database schema changes (use database skill).

Third-party service integrations including Stripe billing, Google Calendar, Twilio (SMS/WhatsApp/Voice), Resend email, Firebase push notifications, analytics (PostHog/GA/TikTok), Langfuse observability, content scraping (Apify/Tavily/Serper), and image generation (Replicate). Use when adding a new third-party integration, modifying billing logic, working with webhooks, configuring analytics tracking, or debugging external API calls. Do NOT use for AI/LLM provider integration (use ai-ml skill) or voice pipeline specifics (use voice skill).

Capacitor v7 iOS/Android native app development. Use when working on native mobile features, Apple/Google Sign-In, push notifications (FCM), haptics, deep linking, safe area handling, Xcode Cloud builds, or platform-conditional code. Covers Capacitor plugin patterns, native bridge communication, and App Store submission. Do NOT use for web-only features or server-side changes.

Real-time communication via SSE broadcasting, NDJSON streaming, WebSocket voice, and in-memory connection tracking. Use when implementing live updates, streaming responses, real-time progress indicators, WebSocket connections, or debugging streaming issues. Covers SSE event patterns, NDJSON response format, connection lifecycle, and client-side EventSource handling. Do NOT use for background job processing (use processor skill) or voice pipeline specifics (use voice skill).

Vue 3 component architecture with shadcn-vue, Reka UI, Tailwind CSS, and design system conventions. Use when building new UI components, modifying existing views, working with the design system, fixing layout/styling issues, or implementing responsive designs. Covers component structure, slot patterns, Tailwind utilities, animation conventions, and accessibility. Do NOT use for state management logic (use state-management skill) or server API routes (use api-server skill).

Voice features including real-time voice mode, speech-to-text (Deepgram), text-to-speech (Fish Audio), voice cloning pipeline, Twilio phone calls, and voice gender classification. Use when working on any voice-related feature, debugging audio issues, or modifying the STT/LLM/TTS pipeline.

React component architecture, dark theme design system, and animation patterns

Claude Opus agentic loop — think, pick tool, execute, repeat

Browser Use Cloud API integration for autonomous web browsing

Convex database schema, mutations, queries, actions, and scheduler patterns

OSINT tool integrations for intelligence gathering

Convex real-time subscriptions powering live investigation updates

Use when the user runs /setup, or asks to set up, initialize, personalize, or re-run setup for this workspace after cloning it. Interviews the user one question at a time, writes their ME.md, scaffolds their project and admin folders, and turns the optional blocks in CLAUDE.md on or off.

Use at the START of any task, before defining or building anything, to decide how the work should be carried out (one live session, a task contract, or a full agent team), how sharply it needs specifying, and whether it genuinely needs file-backed continuity. Session work is ticketless by default. Use when unsure whether a task needs a contract or just a session, or when a raw idea needs to become a well-defined task.

Use when the user wants an agent session supervised while they are not watching it. Triggers on "launch this and watch it", "keep an eye on that session", "drive this to done while I'm out", "what are my sessions waiting on", or any task long enough that the user will not sit through it. Launches a supervised session, reports what is live and what each one is blocked on, and puts a driver on one session that answers routine prompts and escalates every real decision to the user.

Use BEFORE starting any medium or large task, including features, bug fixes, refactors, multi-file changes, branch/commit/push/PR/merge work, and anything touching production or staging, deploys, hosting, databases or migrations, auth, payments, or third-party integration config. Forces a preflight task contract (target, context, acceptance criteria, proof plan, constraints, stop conditions) and an explicit human checkpoint before implementation, so agents never define "done" for themselves.

Interview the user relentlessly about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".

File upload, parsing (PDF, PPTX, DOCX, ODT, HTML, MD), and Convex storage for lesson materials

Convex database schema, queries, mutations, actions, and real-time subscriptions for LessonPlay

Game state machine, lobby system, question flow, scoring, and real-time game coordination

Gemini API integration for generating pedagogically-grounded game questions from lesson content

shadcn/ui patterns, game UI components, responsive layouts for classroom use

Patterns learned and validated by Psi

Project scaffolding — template copy, Claude-powered placeholder filling, git init

Multi-phase loop that runs Claude Code sessions with signal detection and state management

Click-based CLI entry points — psi new and psi run commands

Two-tier learning system — Tier 1 fast capture via LEARNINGS.md, Tier 2 validated promotion to expertise.yaml

Daytona SDK integration for sandbox lifecycle management. Use when working with SandboxManager, creating/destroying sandboxes, executing commands in sandboxes, or transferring files between sandboxes and local filesystem.

xterm.js terminal integration for streaming agent output. Use when implementing AgentCard terminal display, configuring xterm addons, or handling real-time output streaming.

Claude Code authentication in headless Daytona sandboxes. Use when implementing AuthManager, setup wizard, credential validation, or handling OAuth token/API key injection.

Claude Code CLI usage for automated/headless execution in Daytona sandboxes. Use when running Claude Code commands, configuring authentication, or setting up non-interactive mode.

SQLite database schema and operations using better-sqlite3. Use when working with StateManager, persisting runs/rounds/agents, or querying run history.

Use this whenever a user wants to add a new feature or explitly states to research a feature/API or building a plan for a new feature. It iterviews the user for feature details (if not provided), research the best API/service for their needs, confirm choice, then gather all implementation notes for their request and save them as a .claude/plans file.

Automatically deploys to Vercel production, uses Vercel MCP to fetch build logs, analyzes errors, fixes them, and retries until successful deployment. Use when deploying to production or fixing deployment issues.

How to create new skills for this project