paresh-maheshwari/morphe-ai

Morphe AI is a workspace that automates the process of analyzing and patching Android app files, known as APKs. It sets up a pipeline of AI agents, each handling one step of the process: examining an app's metadata, decompiling it into readable code, hunting for specific parts of the code to change, writing a patch, and then building and deploying the modified app. It is built around a tool called Kiro CLI, but the author notes that the instructions and prompts are written so they can be adapted to other AI coding assistants as well. The real content of the project lives in a set of markdown files organized into folders called steering, prompts, and skills. These describe things like how to write patches, how to read a low-level bytecode format called smali, and patterns learned from other community patch repositories, such as bypassing app billing checks or removing ads. Someone using a different AI tool can feed these files in as context or system instructions rather than using Kiro directly. Setting the project up involves installing several Android reverse engineering command line tools, cloning the repository along with a separate repository of your own patches, and creating a configuration file with API tokens for services like Kaggle and GitHub. Kaggle is used specifically because decompiling large apps needs more memory than most personal computers have, so the project runs that step remotely on Kaggle's free notebook servers. The workspace includes six specialized AI agents that each activate for a different task, such as identifying an APK's structure, decompiling it, or building and deploying the final patched file. This project is licensed under a proprietary license, meaning the usual open source freedoms to reuse or redistribute the code do not automatically apply.