Appllama/appllama-skills

A builder, not just a researcher. Agent skills that turn top-grossing app patterns into native-quality mobile screens.

What it solves

It provides AI agents with the specialized knowledge and standards required to build high-quality, native-feeling mobile apps. It bridges the gap between generic AI code generation and the professional design standards of top-grossing mobile applications, ensuring that generated screens are not just functional but meet industry-level UI/UX bars.

How it works

The project consists of two primary "skills" that can be installed into AI agents (like Claude Code or Cursor):

  1. Usage Skill: Acts as a research engine that uses the Appllama MCP (Model Context Protocol) to study real screens, flows, and UI patterns from successful apps to extract design languages.
  2. Design Skill: Enforces a strict "build bar" for Expo and React Native, focusing on Apple HIG fidelity, semantic colors, native controls, and precise navigation and motion logic. It utilizes a full-motion simulator loop to verify flows frame-by-frame rather than relying on static screenshots.

Who it’s for

Developers and designers using AI agents to build or improve mobile applications who want their apps to look and feel like professional, top-grossing products.

Highlights

  • Simulator-Verified Quality: Uses a full-motion simulator loop to ensure animations and flows are polished.
  • Native Fidelity: Focuses on native-feeling Expo/React Native screens with strict adherence to Apple HIG.
  • Research-Driven Design: Integrates with a library of top-grossing apps to study winning patterns before building.
  • Anti-Slop Discipline: Implements strict rules for motion, navigation (e.g., push vs. replace), and thread management to avoid common AI-generated UI mistakes.

Related

  • Project
    deepsense-ai/ragbits

    A modular framework of building blocks for rapidly developing GenAI applications, featuring tools for RAG, multi-agent workflows, and full-stack chat infrastructure.

  • Project
    devenjarvis/lathe

    A tool that uses LLMs to generate personalized, hands-on technical tutorials and provides a local UI for working through them manually to ensure the user actually learns the material.

  • Project
    skyllwt/AutoSci

    A memory-centric AI research agent that automates the full scientific research lifecycle, from literature ingestion and idea generation to experimental execution and manuscript writing.

  • Project
    Gitlawb/zero

    A terminal-based AI coding agent that allows users to edit files and run commands using their choice of LLM provider while maintaining local control and security.