Maestro
A GitHub app that scores a pull request 0 to 100 across behavior, risk, performance, UX, and security, and returns ship, review, or wait.
What it is
Maestro is a GitHub app that analyzes a pull request in real time and produces a single Launch Readiness Score from 0 to 100, with a three way recommendation: ship (80+), review (60 to 79), or wait (below 60).
The problem it solves
Large, AI generated diffs make line by line review ineffective. The useful question shifts from “does this code look right” to “what behavioral, performance, security, and UX changes does this introduce”, so a team can decide to ship or wait without reading thousands of lines.
How it works
When a PR opens, a webhook queues a job that fetches the diff and runs five analyzers in parallel: behavioral (user flow impact via an LLM), security (vulnerabilities and secrets), performance (latency and memory estimates), UX (before and after screenshots via Playwright), and risk (permission changes). A weighted formula combines them into the score, results are stored in PostgreSQL, and a check run plus a PR comment report the score with the blocking issues.
Where it fits
Maestro is a gate between agent generated code and production. It complements Vela: Vela models trust and merge gates across the forge, Maestro produces the per PR readiness signal that such a gate can consume.
Status
The scoring engine, GitHub integration, all five analyzers, and test generation work. The pipeline editor and policy engine are UI shells without execution logic, and there is no production deployment yet.