SuperAPI AI Directory

vibe-coding-ship

Coding & development

View repository

Updated 10-16 03:12

About this skill

Vibe Coding SHIP Phase

Pre-flight → deploy → verify. Routes to the right platform skill.

Prerequisites

  • BUILD phase complete (all phases done, no open debug sessions)
  • User explicitly requests deployment

Step 1 — Platform Detection

Read run_target from progress.txt. Not found → read ./docs/TECH_STACK.md Run Target section. Still unknown → ask once:

Where is this app running?
1. Local machine (terminal)
2. Vercel
3. Netlify
4. DigitalOcean
5. Desktop app (Electron / Tauri)
6. Other — specify

Save answer to progress.txt as run_target.


Step 2 — Pre-Flight Checklist

Run vibe-coding-doc-review — verifies docs match code. Run vibe-coding-security-review — CRITICAL or HIGH findings BLOCK deployment.

Then jump directly to the checklist that matches run_target — do not read all checklists:

Local / Desktop Checklist

PRE-FLIGHT — LOCAL
App starts without errors (npm run dev / python app.py / etc.)
All required env vars in .env / .env.local
Database connection works (migrations run if applicable)
All tests passing
No missing packages (node_modules / venv up to date)

[Python only]
  Migrations run (alembic upgrade head / python manage.py migrate)
  Virtual environment active
  Required services running (Postgres, Redis, etc.)

[Desktop only]
  npm run [electron:build / tauri build] completes
  Output installer/binary exists in dist/ or target/release/bundle/

Cloud Deploy Checklist (Vercel / Netlify / DigitalOcean / other)

PRE-FLIGHT — CLOUD
Production env vars set (no dev values / no hardcoded secrets)
API keys configured for production
All tests passing
No console.log / print debug statements
CORS configured properly
Rate limiting enabled
Build size acceptable

[Python only]
  DEBUG=False in production settings
  SECRET_KEY is a random value (not dev default)
  Migrations run
  Static files collected (python manage.py collectstatic)
  WSGI server configured (gunicorn in Procfile or platform config)

If any item fails: fix it before proceeding. Do not deploy with failing checks.


Step 3 — Deploy

Route to the correct skill based on run_target:

run_target Skill
local vibe-coding-local-runner
vercel vibe-coding-deploy-vercel
netlify vibe-coding-deploy-netlify
digitalocean vibe-coding-deploy-digitalocean
desktop vibe-coding-local-runner (packaging mode)
other_cloud vibe-coding-cli-runner with platform command from run_target_detail

For other_cloud where run_target_detail is Railway / Fly.io / Render / Heroku: warn user that a platform config file is required before deployment:

  • Railway → railway.toml
  • Fly.io → fly.toml (generate with fly launch)
  • Render → render.yaml
  • Heroku → Procfile

Activate the skill and exit. That skill handles all deployment steps.

After deploy skill returns success → confirm smoke test: ask user to verify [deployment_url] loads correctly → if confirmed: write smoke_test: passed to progress.txt. Note that monitoring should be configured via platform dashboard.


Progress Tracking

PHASE: SHIP
STATUS: in_progress
deployment_target: [platform]
started: [timestamp]

CHECKLIST:
✓ Doc review - Complete
✓ Security review - No blockers
✓ Pre-flight checks - Complete
⊙ Deploying - In Progress
✗ Smoke tests - Not Started
✗ Monitoring - Not Started

NEXT ACTION: [current step]

Completion

PHASE: SHIP
STATUS: complete
completed: [timestamp]
deployment_url: [URL if applicable]
deployment_successful: true

PROJECT COMPLETE!