Introduction
Diffowl is a self-hostable pull-request review system for small, AI-accelerated product teams. Review OWL runs an evidence-backed first pass so a person can focus on product judgment and material findings.
The product promise
Section titled “The product promise”For every eligible pull request, Diffowl:
- classifies the execution context before risky work begins;
- loads review policy from the trusted base revision;
- runs reviewer, challenger, and verifier roles;
- normalizes supported claims into material findings or advisory suggestions;
- returns a typed outcome that preserves incomplete, limited, and failed states;
- publishes actionable findings through GitHub when the trust class permits it.
Material findings and advice
Section titled “Material findings and advice”A material finding claims that a change causes a meaningful problem. It includes a location, impact, evidence, lifecycle state, and verification state. Material findings affect review readiness.
An advisory suggestion is a non-blocking improvement. Advice remains separate so style preferences and optional polish do not masquerade as defects.
What Diffowl does not do
Section titled “What Diffowl does not do”Diffowl does not provide:
- a correctness guarantee;
- autonomous merge authority;
- default automated approval;
- a hosted dashboard or required hosted service;
- broad coding-agent behavior or automatic patch application;
- first-class support for GitLab, Bitbucket, or Azure DevOps in V1.