Connecting code review to acceptance criteria

The bottleneck shifted to review

As the team adopted AI-assisted coding, the volume of pull requests increased and review became a bottleneck. Generating changes quickly was creating more work for reviewers to assess.

I built a VS Code extension for an EPAM client to bring acceptance criteria and code changes into the same review workflow, giving developers specific feedback they could act on.

A review workflow inside the editor

The extension was installed on team developers’ machines. An initial setup connected Jira and GitHub using each developer’s credentials.

From chat, a developer could trigger a review through a slash command with a Jira ticket or branch reference. The workflow fetched the acceptance criteria and examined the feature-branch code changes using Copilot agents and skills.

From requirements to actionable findings

The review checked changes against acceptance criteria and looked for potential breaking changes, side effects, and memory leaks.

Feedback identified the relevant lines of code, explained the potential issue, and provided recommendations the developer could act on. The aim was to make findings specific enough to investigate and fix. The extension currently provides feedback only: developers review the findings and decide which recommendations to apply. It does not automatically modify code or post comments to GitHub or Jira.

Observed team impact

After adoption, I observed pull requests moving through approval faster, with clearer feedback and actionable recommendations for developers. This is a qualitative observation; review-time savings have not been quantified.