AI QA for AI-generated code

Is this delta safe to merge?

When an agent writes a PR, Izri runs four independent quality signals against the diff and aggregates them into a single izri/quality verdict. One hard rule fails the delta; a chorus of weak soft signals fails it too.

Four signals, one verdict

Each signal posts its own GitHub Check Run; the umbrella aggregates them into the single status a reviewer or another agent reads.

Where the verdict lands

Same scoring engine, three surfaces. Pick whichever your team already lives in.

GitHub Check Run

One Check Run per signal plus the izri/quality umbrella. Pin the umbrella to branch protection and the rest stay informational, or pin individual children when you want surgical gates.

Action setup

Sticky PR comment

A single comment, edited in place on every push. Shows the umbrella verdict, every contributing signal, hard findings, and AI-generated insights when enabled. Never stacks duplicates.

Scope details

Dashboard report

Per-delta report at /organizations/.../deltas/<id> with line-level annotations, a visual-diff drill-down, a project-level quality-trends chart, and deep links from the sticky comment.

Dashboard docs

Start scoring every delta

Drop the GitHub Action workflow in, point a CLI at your project, or mount the MCP server in your agent. The four signals score the next PR you ship.