DEVLOPN Audit

I added one line to my Definition of Done — /review before pushing

An AI-agent review pass on the diff before even opening the PR — human reviewers finally read architecture and decisions instead of hunting for typos.

I added one line to my Definition of Done.

Before pushing: /review.

One more line in your DoD — before pushing, one review pass by an AI agent: AI re-reads the diff, catches the trifles, human reviews the architecture

The principle

An AI-agent review pass on my diff, right inside Claude Code or Copilot, before even opening the PR. It catches the silly stuff:

  • a forgotten print
  • an unhandled edge case
  • a variable name that lies
  • missing error handling

The result

My human reviewers finally read architecture and decisions, not typos.

It doesn't replace human review. It respects it. You don't hand someone a rough draft to proofread.

And you — does your DoD have an AI pass?