An Autonomous Agent for Completing, Debugging, and Uploading Code to GitHub
CodeAgent is an intelligent autonomous coding agent that takes a coding API/task as input, completes the task, generates the final code in a new file, and automatically pushes the result to GitHub. It can also debug and refactor code as needed using agentic workflows.
This is SukeshOfficial's first Code Agent project, designed to demonstrate intelligent end-to-end software automation.
SukeshOfficial
GitHub: @sukeshofficial
.
├── .vscode/ # VS Code workspace settings
├── .gitignore # Git ignored files list
├── BinarySearch.java # Example task output file
├── README.md # Project documentation
├── agent.py # Core agent logic
├── cli.py # Command-line interface for interacting with the agent
├── git_utils.py # GitHub integration and automation utilities
- ✅ Takes coding task or API as input
- 🧠 Uses an agentic model to solve or debug the problem
- 📄 Generates new output files (e.g.,
.py,.java, etc.) - 🚀 Automatically commits and pushes changes to GitHub
- 🐞 Useful for debugging and automated PR workflows
- Input: User provides a prompt or API instruction (e.g., “Create a binary search in Java”)
- Agent Engine:
agent.pyinterprets the task using a coding LLM - File Generation: Creates the output code in a new file (e.g.,
BinarySearch.java) - Git Automation:
git_utils.pycommits and pushes to the GitHub repository
git clone https://github.com/sukeshofficial/code_agent.git
cd code_agent
# Optional: Create virtual environment
python -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windows
# Install dependencies
pip install -r requirements.txtpython cli.py --task "Create a Python script to calculate factorial using recursion"Output will be saved in a new
.pyfile and automatically pushed to GitHub.
You may use a .env file to store your GitHub token:
GITHUB_TOKEN=your_personal_access_tokenEnsure .env is listed in .gitignore.
MIT License
🚀 Built by SukeshOfficial