Reviewer Assignment¶
This guide explains how CloudAEye suggests who should review a pull request, and what they should check.
Overview¶
CloudAEye suggests a reviewer for an open pull request when you run /cloudaeye:assign-reviewer from your coding agent. It runs from the CloudAEye plugin for Claude Code, not from a PR comment.
It answers the question after the Risk Score: the risk score says whether a person should review the change, and Reviewer Assignment says who. The result names one recommended reviewer with the reasons and their review queue, three to five things to check, and the others worth asking. It only suggests: a review is requested on GitHub only when you ask for one.
Prerequisites¶
Step 1: Register¶
Sign up with CloudAEye SaaS.
Step 2: Setup Code Review¶
Setup Code Review by following the Getting Started guide.
Step 3: Install the Plugin¶
Install the CloudAEye plugin by following Claude Code → Setup.
Suggest a Reviewer¶
Run the following in your coding agent, with the pull request's number:
/cloudaeye:assign-reviewer #412
Add a note in your own words to steer the suggestion:
/cloudaeye:assign-reviewer #412 Tom is away this week, suggest someone else
/cloudaeye:assign-reviewer #412 this is urgent, I don't want someone who is busy
Someone the note rules out leaves the list and is named under Left out with the reason. A note asking for speed puts someone with room in their review queue first.
To request the review, ask for it after the card, for example request a review from Alice. CloudAEye requests it on GitHub through the CloudAEye GitHub App, and only from people the card lists.
Reviewer Card¶
The card includes:
- Recommended reviewer: their name and role, built from code ownership and the teams in
OWNERS.md, never a job title. - Why: the main reason, then up to four measured facts, such as how much of the rewritten code they wrote, and their review queue.
- Suggested review focus: three to five things to check, each naming the file or bug it is about.
- Also consider: the others worth asking, each with a reason and their queue.
- Left out: the author, people who cannot review in the repository, and anyone your note ruled out.
- Based on: what was read, and what was not measured on this commit, with the command that measures it.
See the full example in Skills & Commands.
What It Reads¶
| Signal | Description |
|---|---|
| Code ownership | The OWNERS.md files on the target branch: who writes each area, its lead contributor, and the teams people have added |
| Rewritten code | Who wrote the lines the pull request changes |
| Review history | Who reviewed pull requests on the same files in the last 6 months, and how those pull requests went |
| Bugs | Bugs reported in the changed files in the last 90 days, and who fixed them, from GitHub issues and Jira when the Jira app is installed |
| Review queue | Each person's pending review requests, recent reviews, own open pull requests, and GitHub status |
| Earlier results | The Risk Score, AI Slop Score, and Critical issues from a CloudAEye review of this commit, when they exist |
Code ownership is the minimum: a repository with only OWNERS.md files still gets a recommendation, and each other source adds reasons when it is available.
How It Decides¶
Only people the evidence finds are suggested, ranked by which part of the change matters most, who knows that part, and who has room. Three things always hold:
- Nobody whose GitHub status is Busy is recommended.
- Someone with a long review queue is not recommended first while someone suitable has room.
- When the risk score is High, the AI slop is High or worse, or a review found a Critical issue, a team lead is among the first suggestions.
Code Ownership¶
Reviewer Assignment reads code ownership from OWNERS.md files in the repository. When the target branch has none, it first creates them at the repository root in your working tree and recommends from them. Commit them through a pull request: once merged, every recommendation reads them.
To create or refresh them yourself, run:
/cloudaeye:create-owners
It runs on your machine from git history and writes OWNERS.md files with each area's contributors and lead contributor, plus a blank Teams section for people to fill in. It never commits or pushes. See /cloudaeye:create-owners for the format.
Incomplete Evidence¶
Evidence that was not available is listed as not measured, never treated as a clean result. When the ranking cannot run, the suggestion falls back to the order of the evidence under the same rules, and the card says so.
Reviewer Assignment does not replace a code review. Use Review Code to check for bugs and security vulnerabilities.