golive-skill: ship your AI-built app on your own accounts

Your AI agent can write the app. Putting it live — hosting, DNS, database, email, payments — is still manual work on accounts you own. golive-skill packages that work as an open-source Agent Skill plus a zero-dependency Node CLI, and it runs on your machine instead of on someone else's platform.

What golive-skill is

golive-skill by mikehasa is an Agent Skill with a companion command-line tool. It is written in TypeScript and released under the MIT license. The repository currently shows 1,227 stars and 96 forks. It is not a hosted deploy service: the project describes itself as a way to take your agent-built product live on your own accounts, with no GoLive account, no backend and no telemetry.

Repo: github.com/mikehasa/golive-skill

The five stages: detect → plan → approve → apply → verify

The README lays the workflow out in five stages. First it detects your stack, so it knows what it is working with. Then it plans every change it intends to make. Then it waits for your approval. Only after that does it apply the changes. Finally it verifies that things are actually up.

That ordering is the whole point. The tool does not start creating resources the moment you point it at a repo — it produces a plan you can read first.

What it actually covers

This is the boring half of shipping, which is exactly why it is worth automating: hosting, database, domain, email and payments. All of it is provisioned against accounts you already own rather than through a platform intermediary. The repo's GitHub topics give a sense of the services it targets — cloudflare, neon, netlify, porkbun, godaddy, supabase and vercel are all listed.

Because there is no hosted backend in the loop, your keys stay on your machine. There is no signup step and no telemetry.

Install it and run the first two stages

Grab the repository, point the skill at your local project, then run the first two stages — detect and plan. The exact commands live in the README, but the shape of a first run looks like this:

bash
git clone https://github.com/mikehasa/golive-skill
cd golive-skill

point it at your project directory

then run stage 1 and stage 2 from the README:

detect -> reads your stack

plan -> prints every change before it touches anything

On that first run you get a readout of what it found and a list of the changes it wants to make. Nothing is applied yet.

Why the approval step matters

Most deployment automation asks you to trust it and watch the logs afterwards. golive-skill inverts that: the plan is the deliverable, and approval is a gate. If the plan says it wants to create a database, wire DNS records for your domain, or set up an email provider, you see that before anything changes. If the plan is wrong, you stop there.

This is also what makes it usable for agent-built projects. Your coding agent wrote the application; you still want a human-readable diff of the infrastructure before it goes live.

Repo details

• Owner and repo: mikehasa/golive-skill
• Language: TypeScript
• License: MIT
• Stars: 1,227 · Forks: 96
• Topics include: agent-skill, agent-skills, ai-agents, claude-code, codex, deployment, devops, dns, hosting, infrastructure, database

That topic list is a decent signal of who it is built for — people running Claude Code or Codex style agents who do not want to hand their credentials to a hosted deploy tool.

Quick recap

• golive-skill = open-source Agent Skill + zero-dependency Node CLI for taking a local project live
• Five stages: detect your stack, plan every change, wait for approval, apply, verify it's up
• Covers hosting, database, domain, email and payments — on your own accounts
• No signup, no telemetry, no hosted backend holding your keys
• TypeScript, MIT licensed, 1,227 stars, 96 forks
• Repo: github.com/mikehasa/golive-skill