Post-Deploy Smoke Test

devopsGitHub EventSonnet 4.6Source

What It Does

After each production deploy, hit critical application endpoints, verify HTTP responses, check for JavaScript errors, and confirm key user flows are operational.

Prompt

Claude Code prompt

You are running the post-deploy smoke test. A deployment to production just completed. Your job is to verify the application is healthy before the team considers the deploy done. ## Critical Endpoints to Check Check the following endpoint categories. Replace `BASE_URL` with the production domain from the deployment environment. ### Public Pages (expect HTTP 200) - `GET /` — Homepage - `GET /about` - `GET /pricing` (if applicable) - `GET /blog` or `/articles` (if applicable) - `GET /sitemap.xml` — must return valid XML with at least one `<url>` entry - `GET /robots.txt` — must not block all crawlers on production ### API Endpoints (expect HTTP 200 or appropriate status) - `GET /api/health` or `/api/ping` — health check endpoint; expect `{ "status": "ok" }` or HTTP 200 - `GET /api/version` (if present) — verify the deployed version matches the expected git SHA or version tag ### Auth Endpoints (expect correct redirect or response, NOT 500) - `GET /login` — must return 200, not redirect loop - `GET /dashboard` — must redirect to login (expect 302/307), not return 500 ### Static Assets - Verify the main JS bundle loads (check for a `<script src="/...">` tag in the homepage HTML) - Verify the main CSS bundle loads (check for a `<link rel="stylesheet">` tag) - Confirm no asset URLs return 404 ## Steps 1. **Run HTTP checks** against each endpoint above. For each, record: URL, expected status, actual status, response time (ms), and PASS/FAIL. 2. **Check for server errors in responses.** If any 5xx response is received, immediately mark the deploy as FAILED and escalate. 3. **Inspect the homepage HTML** for obvious render failures: - `<title>` tag must be present and non-empty - `<body>` must not contain a raw error stack trace or "Internal Server Error" text - `__NEXT_DATA__` (or framework equivalent) must be present and parseable JSON 4. **Verify the health check endpoint** returns a successful response within 1000 ms. If it times out or returns an error, mark FAILED. 5. **Check for redirect correctness.** Confirm `http://` redirects to `https://`, and `www.` redirects to the apex domain (or vice versa per site convention). ## Output Format ### Deploy Info - Timestamp - Environment: production - Base URL checked ### Smoke Test Result: PASS / FAIL / DEGRADED ### Endpoint Check Results Table: URL | Expected Status | Actual Status | Response Time | Result ### Issues Found If none: "All checks passed." If any failures: numbered list with URL, failure type, and recommended action. ### Escalation If overall result is FAIL: "ESCALATE: Production is degraded. Notify on-call engineer. Consider rollback." If overall result is DEGRADED: "WARNING: Non-critical issues detected. Review before next deploy."

Trigger

Schedule

Run on a recurring cron schedule

GitHub Event

Run when a GitHub webhook event fires

API

Trigger from your own code by sending a POST request

Event: deployment on main

Quick Start

GitHub event triggers must be configured in the web UI:

Go to claude.ai/code/routines and click New routine
Paste the prompt above
Select GitHub event trigger for deployment events on branch main
Set the model to Sonnet 4.6
Select your target repository and click Create

Manage all your routines at claude.ai/code/routines

Prompt

Claude Code prompt

Quick Start

GitHub event triggers must be configured in the web UI:

Go to claude.ai/code/routines and click New routine
Paste the prompt above
Select GitHub event trigger for deployment events on branch main
Set the model to Sonnet 4.6
Select your target repository and click Create

Manage all your routines at claude.ai/code/routines