GitToSkill

lovable.mp4

GitToSkill turns a GitHub profile into an installable Cursor skill.

Give it a profile like @steipete, and it will:

1. Fetch the profile, profile README, and a notable repo shortlist through GitHub GraphQL.
2. Pull a second batch of repo details to capture style signals from real projects.
3. Use Azure OpenAI to write a profile-specific skill guide.
4. Let you either download the generated markdown skill or install it through the CLI.

Core flow

Web:

• Paste a GitHub profile like @steipete or https://github.com/steipete
• Review the generated skill
• Copy the install command or download the .md file

CLI:

npx gittoskill add @steipete

The CLI calls the GitToSkill backend, writes the returned SKILL.md plus reference files into a local generated folder, and then forwards installation to skills add.

Environment

The generator expects:

• GITHUB_TOKEN
• AZURE_OPENAI_API_KEY
• AZURE_OPENAI_BASE_URL
• AZURE_OPENAI_DEPLOYMENT_NAME_MAP

Optional Azure tuning:

• AZURE_OPENAI_MODEL
• AZURE_OPENAI_REASONING_EFFORT

For local CLI usage against a local dev server, set:

• GITTOSKILL_API_BASE_URL=http://localhost:3000

Otherwise the CLI defaults to the production GitToSkill deployment URL.

Local development

Prerequisites

• Node.js 18+
• pnpm

Install

pnpm install

Run the web app

pnpm dev

Run the CLI locally

$env:GITTOSKILL_API_BASE_URL="http://localhost:3000"
pnpm cli:add -- @steipete

You can forward normal skills add flags after the profile input:

pnpm cli:add -- @steipete --agent cursor --scope project

Project structure

gittoskill/
├── app/                    # Next.js website + profile skill API
├── bin/gittoskill.mjs      # CLI wrapper around backend generation + skills add
├── components/             # Web UI
├── lib/                    # GitHub, Azure, parsing, and skill generation helpers
└── README.md

License

MIT