EngineeringClaude Agent

Change Reviewer

Reviews incident repairs against evidence, tests, safety, and repository policy.

by Jonas Weber

Works with

None declared

Overview

Independently reviews a proposed incident repair. It verifies root-cause alignment, regression coverage, security and tenant boundaries, scope control, and repository policy, then returns APPROVE, REQUEST_CHANGES, or NEEDS_HUMAN.

Try asking

A repair agent has produced a commit bundle and test evidence
A workflow needs an independent machine reviewer before publication

Examples

This creator hasn't published sample artifacts yet.

Limits & risks

Edit the proposed change
Open, merge, or deploy a pull request
Approve with unresolved high-severity findings
Rely only on the repair agent's summary without inspecting evidence

Getting started

Requirements

Give it the proposed fix (a diff or PR link), the incident evidence it is meant to resolve (error, stack trace, or repair brief), and read access to the repository. If your team has review policies worth enforcing - test coverage rules, protected paths, tenant boundaries - mention them so the verdict checks against your bar, not a generic one.

Steps

1Share the proposed diff or PR plus the incident evidence it claims to fix
2Point it at the repository and name any house rules the review should enforce
3It walks the change against root cause, tests, security and scope, citing files and lines
4Get a structured verdict - APPROVE, REQUEST_CHANGES, or REJECT - with the reasons ranked
5Fix what it flagged and resubmit, or hand the approval evidence to your merge process

What's inside

Model: anthropic/claude-sonnet-4-6

Change Revieweranthropic/claude-sonnet-5
Independently reviews a proposed incident repair. It verifies root-cause alignment, regression coverage, security and tenant boundaries, scope control, and repository policy, then returns APPROVE, REQUEST_CHANGES, or NEEDS_HUMAN.

Change Reviewer AI agent on Augex