devlearn-post-ship

name	devlearn-post-ship
description	Teaches post-deploy verification: smoke tests, health checks, monitoring vocabulary, rollback, and declaring a release healthy. Use after deploy, "is prod ok", "post ship", canary, or when user shares a live URL. Pairs with devlearn-deploy and devlearn-pre-ship. Voice triggers: "post deploy", "prod check", "smoke test", "monitoring", "is it live".

DevLearn: Post-Ship

Iron law

Verify in production-like conditions. Incognito, real HTTPS URL, critical path once — not only on your logged-in laptop.

Voice

Follow ../shared/voice.md. Celebrate working URL — then teach what to watch.

Context

Deploy gets code live. Post-ship proves real users won't hit a broken path — and teaches what to do when something fails (rollback, logs, smoke tests).

Lifecycle: ../shared/ship-lifecycle.md

Prerequisites

• Live URL (prefer HTTPS) or staging URL treated as prod drill
• Know what changed this release (commit, PR, or explain-diff)

Before you start

1. Parse DEVLEARN.md lifecycle.post_ship_verify
2. Capture URL, commit SHA, deploy time if available
3. If no URL yet → suggest /devlearn-deploy first

Phase 1: Capture deploy facts

**Live URL:** https://...
**Commit/version:** abc123 / v1.2.0
**Release notes:** [one line user outcome]
**Changed areas:** [from diff summary]

Link /devlearn-explain-diff if user unsure what shipped.

Phase 2: Smoke checklist

Run reference.md checklist — document each row:

Check	Why teach
Homepage loads incognito	Cached auth/cookies hide bugs
Critical user path	The job your app exists for
API health or equivalent	Server actually up
Wrong env symptoms	Blank data, 500 on API only in prod
Second network / phone	Not just your WiFi
HTTPS redirect	Users hit secure URL

Viber: top 3 checks + term smoke test.

Seasoned: add latency spot-check, error budget mention if relevant.

Write results to .devlearn/post-ship-log.md from ../.devlearn/post-ship-log-template.md when progress: true.

Phase 3: Monitoring basics (curious/deep)

Concept	Plain English
logs	server diary of errors
metrics	numbers over time (latency, error rate)
alert	notify human when threshold crossed
canary	small % traffic to new version first
rollback	revert to last good deploy
SLO	target reliability you aim for (seasoned)

Don't setup full observability unless asked — teach vocabulary + one practical log location on their host.

Phase 4: When something's wrong

1. Rollback first if users blocked — teach platform-specific path (reference.md)
2. Then debug with /devlearn-debugging using prod symptoms
3. Seasoned postmortem (optional): which pre-ship check would have caught it → update checklist

Never debug prod by guessing — reproduce in incognito, read logs, compare env to staging.

Phase 5: Declare healthy

## Post-ship status

**URL:** https://...
**Smoke tests:** pass / fail [list failures]
**Watch until:** [date + 24h default for risky releases]
**Rollback:** [one command or dashboard path]

### Term of the moment
**[smoke test]**: quick check that critical paths work where users actually go

Persona integration

Persona	Emphasis
viber	Incognito + one happy path + celebrate
seasoned	Rollback command, log query, watch window
autodetect	Match release risk

Handoffs

Situation	Suggest
Errors in prod logs	/devlearn-debugging
Need automated smoke in CI	/devlearn-devops
Next release	/devlearn-pre-ship
Session wrap	/devlearn-recap
Security incident	/devlearn-security

Common mistakes

Smell	Fix
Test only while logged in as dev	Incognito path
"Homepage works" only	Run critical business path
Ignore API 500 on prod	Network tab on same URL users use

STOP checkpoint

If smoke fail:

"Rollback now (users broken) or hotfix forward? I'll teach rollback steps either way."

Required footer

---
DevLearn status: DONE | BLOCKED
Production healthy: yes | no | monitoring
Smoke passed: N/M
Suggested next: /devlearn-recap | rollback | /devlearn-debugging | /devlearn-devops
---

Additional resources

• Checklist detail → reference.md
• Example log entry → examples.md
• Deploy → devlearn-deploy