DEPLOY Framework — Ship It Right the First Time

Dev Trilogy

CONSTRUCT

→

DEPLOY

→

DIAGNOSE

Document the State

Capture exactly what you built, what it depends on, and what environment it needs. If it can't be rebuilt from notes, it's not ready to ship.

Environment Audit

Confirm the production environment matches your assumptions. Port conflicts, dependency versions, and permission issues get caught here, not at 2am.

Pre-flight Checklist

Run through your launch sequence before touching production. Secrets loaded, startup verified, rollback path confirmed.

Launch Sequence

Execute the deployment in the documented order. No improvising. The sequence exists precisely because production is not the time for creative problem solving.

Observe & Verify

Don't walk away after deployment. Watch logs for 10 minutes. Test the primary use case through the production environment. Confirm real traffic, not just startup.

You're Done. Now Document It.

The deployment is live. Write the post-deploy note while it's fresh: what worked, what you'd change, what the next Claude Code session needs to know.

URL confirmed live

Primary feature tested

Rollback path clear

Deploy note written

MEMORY.md updated

The middle of the trilogy

DEPLOY sits between CONSTRUCT (building it) and DIAGNOSE (fixing it). Most deployment failures happen because developers skip directly from "it works" to "it's live" without the disciplined verification steps in between.

Phase 5, "Observe and Verify," is the one most people skip. They deploy, see no immediate error, and walk away. But production bugs often take minutes to surface: the first real user hits an edge case, a queue fills up, a rate limit kicks in. Ten minutes of watching logs after deployment catches problems that would otherwise become incidents.

The interactive checklist on Phase 6 is there for a reason. Every item represents a deployment that failed because someone forgot that specific step. The checklist is the cheapest insurance policy in software.