Step 6 Deploy Guide

Step 6 · Deploy Guide

Command: /deploy-guide [staging|production] · Habit: H1 Be Proactive · Previous: Step 5 · Review AI · Next: Step 7 · Monitor Setup

Purpose

Produce a staging-first deploy plan with a rollback ready to paste. Prevents the class of incidents caused by "let's just push it, it looks fine."

When to use

• Any deploy touching production
• Any database migration
• Any config change to shared infrastructure
• Any production canary or provider-managed capacity change, such as nodegroup/ASG scale-down

When to skip

• Never for production. Local-only changes do not need this step.

Process

1. Read existing deploy scripts, CI/CD configs, docker-compose.yml, Makefile
2. Produce staging steps first, with exact commands
3. Define staging verification — health check, smoke test, feature verify
4. Produce production steps (only after staging QA passes)
5. Produce rollback procedure — copy-pasteable commands
6. For provider-managed canaries, add reconciliation gates: planned target vs actual provider-selected target, desired/min/max alignment, schedulable capacity, and no unintended SchedulingDisabled nodes
7. Produce post-deploy verification

Output

A deploy plan document containing:

• Deploy: [version/feature] title
• Staging — numbered command list
• Staging QA — verification checklist (health, feature, smoke test)
• Production — numbered command list (only after staging QA passes)
• Rollback (ready to paste) — copy-pasteable commands inside a fenced block
• Canary reconciliation — planned target, actual provider target, scale/update status, scheduling state, and mitigation path
• Post-deploy verification — checklist

Rules

• Never combine build + deploy in one step for production
• Never skip staging
• Never deploy on a Friday afternoon without explicit reason
• Never call a provider scale-down complete until Kubernetes scheduling state is reconciled

Handoff

• Expects: Reviewed, passing code from /review-ai
• Produces for /monitor-setup: A deployed service ready to observe

H1 Checkpoint

Important

"Am I preventing future incidents, or reacting to current ones?"

See also

• Source: skills/deploy-guide/SKILL.md
• Habits Reference → H1