review-step

What problem does it solve?

Manually invoke a reviewer agent to validate the work produced for a specific epic and step when the normal build-step review flow is not appropriate or when an automated review must be re-run.

Core Features & Use Cases

• Determine context: Locate the active epic and step from execution state or accept explicit arguments to target a specific execution.
• Change detection: Compute the git diff since the last approved step or TDD baseline to scope the review.
• Reviewer orchestration: Dispatch the reviewer subagent with the Task tool, providing the step specification, diff, test results, and review criteria, then process the review output to mark approval or surface requested changes.
• Use Case: When a developer iterates on a model step outside the automated pipeline, run this Skill to get a formal reviewer verdict and update execution state accordingly.

Quick Start

Run the review-step skill to invoke the reviewer agent for the current epic and step, collect the git diff and test results, and produce an approval or requested-changes response.