What problem does it solve?
It turns a natural-language request into a complete, validated, and published Android plugin app, eliminating the manual work of scaffolding HTML frontends, defining agent-callable tools, wiring connectors, and verifying data consistency before release.
Core Features & Use Cases
- End-to-end project scaffolding: Generates a standalone phone-first HTML/CSS/JavaScript app with a hand-authored SVG icon, a Xiaowan SKILL.md, and a toolkit.json declaring agent-callable tools bound to sqlite, xiaowan, or http_json connectors.
- Validation and publishing pipeline: Runs project_check to audit Data/Tool/Display consistency and no-mock rules, then project_publish to install the plugin and add it to the Android home screen.
- Real-data enforcement: Requires every business field to come from user input, a declared connector, or runtime AI generation, rejecting placeholder arrays, canned AI answers, and fabricated runtime status.
- Use Case: Ask the assistant to "create a weekly workout coach app" and it produces the dashboard, SQLite schema, AI plan-generation tool, validates a create/read round trip, and publishes the plugin to the home screen.
Quick Start
Ask the assistant to create an app for your idea, for example: build a habit tracker app with AI-generated weekly coaching plans and publish it to my home screen.