Skynet

Skynet

Overview

Skynet is an AI-powered coding assistant inspired by tools such as OpenCode and Codex. It allows developers to interact with their codebase through natural language, inspect project files, generate or modify code, and complete development tasks from a single interface.

The Challenge

Traditional AI chat tools can generate code, but they often lack awareness of the wider project structure. Building a useful coding agent requires the system to understand relevant files, decide what needs to change, use development tools, and apply edits without forcing the developer to manually copy code between applications.

Solution

I built Skynet as an agent-based development application that connects large language models with a local codebase. The application can inspect project files, understand coding requests, determine which files are relevant, and perform code-generation or editing tasks through an iterative agent workflow.

The platform is model-independent, allowing different AI models to be selected based on the task, cost, speed, or reasoning capability required.

Key Features

• Natural-language interaction with a codebase
• AI-assisted code generation and editing
• Project file exploration and contextual understanding
• Automatic identification of relevant files
• Support for switching between multiple AI models
• Tool-based agent execution
• Multi-step coding workflows
• Codebase-aware conversations
• File creation and modification
• Developer-focused terminal-style interface

Technical Highlights

• Built an agent loop that processes requests, selects tools, reads results, and continues until the task is complete
• Implemented codebase context retrieval for more accurate responses
• Developed file-system tools for reading, creating, and updating project files
• Added support for configurable language-model providers
• Managed conversation and tool-call state across multi-step tasks
• Designed safeguards around file edits and tool execution
• Built the application with a modular architecture so additional tools and models can be added easily

Tech Stack

React • TypeScript • Node.js • Large Language Models • Tool Calling • File-System APIs • AI Agent Architecture

My Contributions

• Designed and developed the core AI coding-agent workflow
• Implemented project file discovery and contextual code understanding
• Built tools for reading, creating, and editing files
• Added support for switching between different AI models
• Developed the agent loop for multi-step task execution
• Built the developer interface and conversation experience
• Continued extending the platform with new tools and capabilities

Outcome

Skynet demonstrates how large language models can be transformed from simple chatbots into practical software-development agents. The project strengthened my understanding of agentic systems, tool calling, codebase retrieval, context management, model integration, and AI-assisted developer tooling.