The Problem

Mobile app development teams struggle to translate high‑level research into production‑ready screens. They need a reliable way to pull design patterns from top‑grossing apps, convert them into native‑quality components, and validate the result in a simulator before committing code.

What This Does

appllama-skills bundles two agent‑ready skills:

  • appllama-usage – a research engine that talks to the Appllama MCP. Its logic lives in skills/appllama-usage/agents/openai.yaml and the reference docs in skills/appllama-usage/references/*.
  • appllama-app-design-skill – a build bar that turns the research output into Expo/React Native screens. The build rules are described in skills/appllama-app-design-skill/SKILL.md and the supporting reference files (native-controls.md, motion.md, etc.).

Both skills are packaged as plain directories so any agent that understands the Skills framework can import them.

How It Is Wired

  • Entry Point – The Skills runtime looks for a plugin.json in the root of a cloned repository. ./.claude-plugin/plugin.json and ./.cursor-plugin/plugin.json expose the two skills to the agent ecosystem.
  • Skill Registration – Each skill directory contains a SKILL.md that the agent uses as a manifest. The file lists goals, input/output contracts, and the associated YAML/JSON files that actually drive the skill.
  • Runtime Dependencies – The skills rely on the external Appllama MCP. skills/appllama-usage/agents/openai.yaml defines the prompt templates and the OpenAI model used. skills/appllama-app-design-skill contains only documentation; the actual generation is performed by the agent calling the MCP.
  • Side Effects – No direct file or network side‑effects are coded here; all interaction occurs indirectly through the agent’s execution context.

How To Use It

  1. Install ``bash npx skills@latest add appllama/appllama-skills ` To add only a single skill: `bash npx skills@latest add appllama/appllama-skills --skill appllama-app-design-skill ``
  2. Configure – The MCP credentials must be set in the agent’s environment (e.g., OPENAI_API_KEY). No additional config files are required in this repo.
  3. Run – Once installed, invoke the skill via the agent’s command interface, e.g., claude-code use appllama-usage. The agent will read the openai.yaml prompts and send them to the MCP.

Real‑World Use

A product manager wants a new onboarding screen.

  1. Agent runs appllama-usage to query the MCP for patterns in successful onboarding flows.
  2. The agent passes the retrieved data to appllama-app-design-skill, which outputs a React Native component that adheres to Apple HIG and native controls.
  3. The component is dropped into the project and verified in the simulator loop described in the skill docs.

Code Health & Issues

  • Medium – No tests – 0 test files detected; code paths are unverified.
  • Medium – No CI – No .github/ or other CI config; automated build/test gates are absent.
  • Low – Minimal documentation – Only two README‑style docs; usage examples are missing.
  • Low – No external secrets – No API keys committed; good from a security standpoint.

The Bottom Line

appllama-skills provides a lightweight, declarative way to bridge research and implementation for mobile UI design. It is ideal for teams that already use the Appllama MCP and want a plug‑and‑play skill set. The repository lacks automated tests and CI, so contributors should add those before shipping in production environments.