alistaircroll

verify-deployment

0
0
# Install this skill:
npx skills add alistaircroll/verbose-deployment --skill "verify-deployment"

Install specific skill from multi-skill repository

# Description

Confirm the deployment succeeded on the hosting platform by polling programmatically β€” no sleep timers. Detect the platform (Vercel, Netlify, GitHub Pages, AWS, etc.) and poll until READY or ERROR. Use as Phase 8 of the deployment pipeline.

# SKILL.md


name: verify-deployment
description: "Confirm the deployment succeeded on the hosting platform by polling programmatically β€” no sleep timers. Detect the platform (Vercel, Netlify, GitHub Pages, AWS, etc.) and poll until READY or ERROR. Use as Phase 8 of the deployment pipeline."


Verify Deployment

Confirm the deployment succeeded on the hosting platform. Do NOT use sleep timers β€” poll programmatically.

Detect

Identify the hosting platform from project config, git remotes, or documentation:

Platform Detection Verification Method
Vercel vercel.json, Vercel project link vercel ls CLI or Vercel API β€” poll deployment state
Netlify netlify.toml, .netlify/ netlify status or Netlify API
GitHub Pages gh-pages branch, Actions workflow gh run list β€” check workflow status
AWS CloudFormation, CDK, SAM config AWS CLI β€” check stack/deployment status
GCP app.yaml, Cloud Run config gcloud CLI
Self-hosted Deployment scripts curl the health endpoint

Poll Logic

Poll the platform's API or CLI for deployment state. Typical state progression:

QUEUED -> INITIALIZING -> BUILDING -> READY
                                   -> ERROR
                                   -> CANCELED
  • QUEUED / INITIALIZING / BUILDING β€” not ready yet, poll again
  • READY β€” deployment succeeded, proceed to production verification
  • ERROR β€” pull build logs, diagnose, fix locally, re-push
  • CANCELED β€” investigate why, may need manual redeploy

Never declare "deployed" based on push alone β€” always confirm READY state via the platform API.

Stuck deployments: If a deployment stays in QUEUED/INITIALIZING for more than 5 minutes with zero build logs, it may be stuck. Check for queue congestion (multiple pending deployments) and consider triggering a fresh deploy.

Collect

  • Deployment ID / URL
  • State transitions with timestamps
  • Build duration
  • Any build errors or warnings

Stop Conditions

STOP if: Deployment fails. Pull build logs, diagnose, fix, restart from Phase 1 (the project-inventory skill).

If no hosting platform is detected: Skip this phase, mark as N/A in the report.

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.