Building copilot-bootstrap: Scaffolding Projects with AI

Overview

Starting a new project from scratch means wiring up the same tedious setup over and over — ESLint configs, tsconfig, CI workflows, README skeletons, Docker files. copilot-bootstrap automates that ceremony by integrating GitHub Copilot at the scaffolding step rather than the editing step.

Motivation

I found myself copy-pasting boilerplate between projects and gradually diverging on config conventions. I wanted a single entry-point that would ask a few high-level questions — language, framework, deployment target — and generate a consistent, opinionated baseline in seconds.

Approach

The CLI is built in TypeScript and runs as a Node.js executable. At startup it:

1. Prompts the user for project metadata (name, description, stack).
2. Calls the GitHub Copilot API to generate config files tailored to the stated stack.
3. Writes the generated files to disk and performs a final pass to resolve any cross-file inconsistencies.

// Example: invoking the scaffold step
const result = await scaffolder.run({
  name: "my-service",
  language: "typescript",
  framework: "express",
  ci: "github-actions",
});
console.log(`Generated ${result.files.length} files in ${result.dir}`);

Tech Stack

• TypeScript for type-safe CLI internals
• Node.js for file I/O and process management
• GitHub Copilot API for content generation
• Commander.js for argument parsing

Outcomes

The tool reduced my per-project setup time from ~30 minutes to under 2 minutes, and the generated configs are immediately usable without manual edits for the common cases.

Conclusion

copilot-bootstrap is a good example of using AI as a development multiplier at the infrastructure level rather than the code-writing level. The repo is open and PRs are welcome.