deployment-plan

Installation
SKILL.md

Deployment Plan

Two phases: grill first, then write. A deployment runbook is only useful if it reflects the real infrastructure and the operator can paste its commands verbatim and have them work. So extract the truth before writing a line of the document.

Running project (adopt mode). If the system already deploys, document how it deploys today. Record each gap you find (no backups, no rollback path, secrets in the repo) as a finding in the runbook and in your report. Do not redesign the deployment here.

Phase 1: Grill

Interview the user relentlessly about how this system actually deploys, one question at a time, walking each branch until resolved. For every question, give your recommended answer and explain the trade-off, but don't move on until they confirm.

Inspect before asking. Read the repo first: docker-compose*.yml, Dockerfile, .github/workflows/, deploy/, nginx configs, .env.example, migration setup, README, any existing CLAUDE.md or AGENTS.md, and in docs/technical-specs/ the security section (the reset-API decision) and the environment configuration (the gating flag names). Every fact you can read, read; don't ask what the repo already answers. Ask only to fill genuine gaps and to confirm inferences.

Cover, at minimum (skip what's irrelevant to this stack, probe what's load-bearing):

Installs
32
GitHub Stars
9
First Seen
Jun 22, 2026