Code review
Explyt separates reviewing changes, independent review, and vulnerability search. These features solve different problems and can be used one after another before a commit or a pull request.
What to choose
| If you need to | Use |
|---|---|
| Review the files the agent changed in the current chat | Agent Changes |
| Have a separate agent review the changes independently | Auto Review |
| Go from an issue in a published report to the line of code and start a fix | Review window |
| Scan a JVM project for vulnerabilities | Vulnerability search |
Recommended sequence
- Review the diff in Agent Changes and reject unrelated edits.
- Run the related tests or the build through an IDE configuration.
- Run Auto Review if you need an independent check: the Review subagent reports its findings back to the main agent in the current chat.
- If the regular review flow published a report to the Review window, work through the findings there and fix the confirmed issues.
- Before a release or an audit, run vulnerability search if needed.
Auto Review checks changes with the Review subagent and reports the results in the current chat. The Review window is for reports that the regular review flow explicitly publishes there. Vulnerability search is a separate analysis of the whole supported project; its results can be attached to a chat for an explanation, a validity check, or a fix.
Practical tasks
Do not accept a change based on the agent's message alone. A verifiable result is a reviewed diff plus tests, a build, or another suitable check that actually ran.