Skip to content

Workflow Overview

github-actions[bot] edited this page Jun 6, 2026 · 8 revisions

Workflow Overview

The workflow is a sequence of markdown skills that keeps AI-assisted work reviewable. Each step has a purpose, a handoff, and a skip rule.

Steps

Step Skill Habit Output
0 /research H5 Research brief
1 /requirements H2 PRD-style scope and success criteria
2 /design H8 Architecture options and decision
3 /breakdown H3 Atomic task list
4 /build-brief H5 Implementation brief
5 /review-ai H4 Review findings before commit
6 /deploy-guide H1 Deploy, rollback, and verification plan
7 /monitor-setup H7 Observability and runbook updates
/research
  -> /requirements
  -> /design
  -> /breakdown
  -> /build-brief
  -> /review-ai
  -> /deploy-guide
  -> /monitor-setup

Use This When

  • The change is larger than a trivial local edit.
  • The domain is unclear.
  • Architecture, data, security, or deployment behavior may change.
  • A production or operational finding needs controlled handling.

Practical Skip Guide

Change type Typical path
Formatting only /review-ai
Obvious one-line bug fix /review-ai
Dependency or config change /review-ai -> /deploy-guide
Small feature /requirements -> /breakdown -> /build-brief -> /review-ai
Medium feature /requirements -> /design -> /breakdown -> /build-brief -> /review-ai -> /deploy-guide
Unclear domain or architecture change Full workflow
Production incident or config hotfix /diagnose or /operational-state, then /consistency-check, /review-ai, /deploy-guide as applicable

Warning

Do not skip /review-ai for AI-generated work. Other steps have legitimate skip cases; review is the baseline.

Operational Extensions

The workflow now includes explicit operational support outside the linear steps:

  • /operational-state classifies findings before mutation or closure.
  • /consistency-check can check incident/config hotfix claims against evidence.
  • /deploy-guide includes reconciliation gates for provider-managed canaries and capacity changes.

See Also

Clone this wiki locally