What are the prerequisites for deployment?

Ensure a Dockerfile exists in the project root and the app listens on port 8080. The deployment flow expects these to be true before attempting deployment.

How do I set up a production remote?

First check for an existing remote with git remote -v | grep production. If none, list servers with bash ${CLAUDE_PLUGIN_ROOT}/scripts/server.sh list, discover domains with bash ${CLAUDE_PLUGIN_ROOT}/scripts/dns.sh list-zones, pick a server and app name, add the remote using git remote add production deploy@ . : . , then configure DNS and push to production main.

How do I troubleshoot a failed deployment?

If deployment fails, SSH into the server as deploy and review the last 20 log entries to identify errors. You can also verify the remote URL, ensure Dockerfile exists, and confirm the app is listening on port 8080. Use deploy.sh info to show remote details.

app-deployment

Scanned

npx machina-cli add skill The-Focus-AI/marina-skill/app-deployment --openclaw

Files (1)

SKILL.md

3.0 KB

App Deployment Skill

You help deploy applications using a git-push-to-deploy workflow.

Setup

Run bash ${CLAUDE_PLUGIN_ROOT}/scripts/check-deps.sh to verify tools are available.

How Deployment Works

User runs git push production main
SSH connects to server as deploy user
SSH forced-command routes to deployer script
deployer creates a bare git repo (first time), receives the push
post-receive hook checks out code and runs docker build
deployer stops old container and starts new one with Caddy labels
Caddy automatically provisions HTTPS and reverse proxies to port 8080

App requirements:

Must have a Dockerfile in the project root
Container must listen on port 8080

Scripts

Deployment utilities use ${CLAUDE_PLUGIN_ROOT}/scripts/deploy.sh:

Copy env to server

bash ${CLAUDE_PLUGIN_ROOT}/scripts/deploy.sh env

Copies .env.production (preferred) or .env to the server and restarts the container.

Restart container

bash ${CLAUDE_PLUGIN_ROOT}/scripts/deploy.sh restart

Force rebuild

bash ${CLAUDE_PLUGIN_ROOT}/scripts/deploy.sh push-empty

Creates an empty commit and pushes to trigger a full rebuild.

Show remote info

bash ${CLAUDE_PLUGIN_ROOT}/scripts/deploy.sh info

Setting Up a New Deployment

Walk the user through these steps:

Check prerequisites:
- Verify Dockerfile exists
- Check the app listens on port 8080
Check for existing remote: git remote -v | grep production
If no production remote, guide setup: a. List servers: bash ${CLAUDE_PLUGIN_ROOT}/scripts/server.sh list b. Discover domains: bash ${CLAUDE_PLUGIN_ROOT}/scripts/dns.sh list-zones c. Pick/create server and choose app name d. Add remote: git remote add production deploy@<server>.<domain>:<appname>.<domain> e. Add DNS: bash ${CLAUDE_PLUGIN_ROOT}/scripts/dns.sh add <appname>.<domain> <server_ip> f. Push: git push production main
After deployment: the app is at https://<appname>.<domain> (Caddy handles HTTPS).

Git Remote Format

deploy@<server-hostname>:<site-fqdn>

Example: deploy@web1.example.com:myapp.example.com

The site FQDN becomes both the Docker container name and the hostname Caddy uses for routing/HTTPS.

Behavior

Check for Dockerfile before attempting deployment
When setting up, show the full git remote URL and confirm
After deployment, tell the user the app URL
If deployment fails, suggest checking logs: ssh deploy@<server> (shows last 20 log entries)
Warn about sensitive values before copying env files

Source

git clone https://github.com/The-Focus-AI/marina-skill/blob/main/skills/app-deployment/SKILL.mdView on GitHub

Overview

Deploy apps using a git-push-to-deploy workflow. It sets up a production remote, pushes code, manages environment variables, and restarts containers, with Caddy provisioning HTTPS and routing to port 8080.

How This Skill Works

On push to production main, an SSH login as deploy triggers a deployer script via a forced command. The script creates a bare git repo if needed, receives the push, and the post-receive hook checks out code and runs docker build, then the deployer stops the old container and starts a new one labeled for Caddy. Caddy then provisions HTTPS and reverse proxies to port 8080.

When to Use It

Deploy a new app to a remote server using a production remote
Update an existing app by pushing code to production main
Copy environment variables to the server before restarting the container
Force a full rebuild after env or dependency changes
Verify the app URL and HTTPS via Caddy after deployment

Quick Start

Step 1: Check prerequisites: verify Dockerfile exists and the app listens on port 8080
Step 2: Check for a production remote; if absent, run server and DNS discovery commands and add the remote, then push to production main
Step 3: After deployment, verify the app at https://<appname>.<domain> and use deploy.sh info to view remote details

Best Practices

Ensure a Dockerfile exists in the project root
Confirm the app listens on port 8080
Check for an existing production remote before pushing
Copy env files with caution and restart the container after deploy
Use the deploy.sh scripts (env, restart, push-empty, info) for common tasks

Example Use Cases

Deploying a new app to a server via git push production main
Updating an existing app by pushing new code to production main and watching deployments
Copying .env.production to the server and restarting the container
Forcing a rebuild with push-empty to trigger a full docker build
Checking remote info after setting up a new deployment

Frequently Asked Questions

Add this skill to your agents