Code Review Settings¶
Overview¶
CloudAEye allows you to adjust Code Review speed and severity sensitivity. This page outlines how to configure these settings.
Prerequisites¶
Setup Code Review¶
Setup Code Review by following the Getting Started guide.
Code Review Configuration Settings¶

Configure¶
- No Secrets in Code – When enabled, the code review will scan for hard-coded secrets, API keys, or credentials.
Select your preferred option and click Save to save your changes.
Ultrareview Settings¶
Ultrareview settings control how CloudAEye performs in-depth code reviews, evaluates pull request risk, determines whether a change can be automatically approved, and assigns human reviewers when additional review is needed.
You can configure default settings at the tenant level. Repository-level settings defined in .cloudaeye.config override tenant-level defaults.
1. Checks¶
Configure which checks CloudAEye runs during Ultrareview.
A disabled check is skipped entirely. For automatic approval, the Bugs, Security, AI Slop Score, and Risk Score checks must all be enabled. If any of these checks is disabled, CloudAEye will not automatically approve changes.

| Setting | Default | Description |
|---|---|---|
| PR Checklist | Enabled | Runs the PR hygiene checklist to identify missing descriptions, incomplete documentation, missing tests, and other pull request issues. Applies to pull requests only. |
| Bugs | Enabled | Analyzes code changes to identify potential bugs, logic errors, and unintended behavior. |
| Security | Enabled | Analyzes code changes for security vulnerabilities, including leaked secrets and other security issues. |
| AI Slop Score | Enabled | Evaluates AI-generated code quality and estimates how much careful manual review the changes require. |
| Risk Score | Enabled | Assesses the overall risk of a change, determining whether human review is necessary and explaining the reasons. |
| Architecture | Enabled | Evaluates the architectural impact of code changes, including adherence to architectural principles and layer boundaries. |
| Control Flow | Enabled | Analyzes changes to execution paths, function calls, and the order in which operations are performed. |
| Task Check | Enabled | Validates code changes against the associated ticket or task to identify missing requirements or implementation gaps. Runs only when a task is provided. |
2. Hard Limits¶
Hard limits define the maximum acceptable risk and review severity for automatic approval.
A change that exceeds any configured hard limit is never automatically approved, even if other checks pass.

Leave a limit empty to let CloudAEye determine the appropriate threshold automatically.
| Setting | Default | Description |
|---|---|---|
| Risk Score | CloudAEye default | Maximum risk score (0–100) a change can have and still qualify for automatic approval. |
| AI Slop Score | CloudAEye default | Maximum AI Slop Score (0–100) a change can have and still qualify for automatic approval. |
| Review Severity | CloudAEye default | Highest finding severity allowed for automatic approval. Supported values are Low and Medium. |
For example, setting the Risk Score limit to 30 means that changes with a risk score above 30 cannot be automatically approved.
Similarly, setting Review Severity to Low means that any finding with Medium or High severity prevents automatic approval.
3. Human Review Policy¶
Configure when and how CloudAEye assigns human reviewers to pull requests that require additional review.
Auto-assign Human Reviewer¶
Default: Enabled
When a change is not eligible for automatic approval, CloudAEye automatically assigns human reviewers based on the configured review policy.
Disable this setting if you want CloudAEye to identify changes requiring human review without assigning reviewers automatically.
Maximum Human Reviewers to Assign¶
Default: 3
Specifies the maximum number of human reviewers CloudAEye can assign to a pull request.
Supported values: 1, 2, or 3.
CloudAEye assigns one reviewer by default. Additional reviewers can be requested through the Instructions setting, up to the configured maximum.

Instructions¶
Provide additional review requirements in plain language. These instructions allow you to define repository-specific or organizational review policies that require additional human oversight.
Instructions can request more human review but cannot override CloudAEye's risk assessment or reduce the required level of review.
Examples:
Changes under services/payments/ always need two human reviewers.Always assign at least one human reviewer, even when CloudAEye approves the change automatically.Changes involving authentication or authorization require two human reviewers.
CloudAEye interprets these instructions when determining the appropriate human review requirements.
Repository-Level Configuration¶
Tenant-level settings provide defaults for all repositories. Individual repositories can override these defaults using a .cloudaeye.config file.
When a repository defines its own Ultrareview settings, those values take precedence over the corresponding tenant-level settings for that repository.
This allows organizations to maintain consistent review policies while accommodating different requirements across repositories.
Configuration Precedence¶
- Repository-level
.cloudaeye.configsettings - Tenant-level Ultrareview settings
- CloudAEye defaults, where no explicit value is configured
Review Settings in Your Repository¶
The severity floor, the paths never reviewed and the review rules that the review commands in
your coding agent use are set in the repository's .cloudaeye.config and .cloudaeye.rules. See
Settings Reference: Code Review.