[BUG] Ultrareview crashes repeatedly on large PRs – no findings produced

Resolved 💬 3 comments Opened Apr 24, 2026 by Passioneers Closed May 30, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Ultrareview crashes on large PRs before producing any findings.
Two consecutive Ultrareviews for the same PR (Passioneers/alice_mvp#24)
both ended with the error "Review crashed before producing findings.
See session logs for details." – no findings were returned in either case.
Both Free Ultrareview credits (2 of 3) were consumed without producing output.

What Should Happen?

The Ultrareview should either complete and return findings, or fail gracefully
with a meaningful error message explaining why it could not finish (e.g. PR too
large, timeout, rate limit) so the user knows how to proceed.

Error Messages/Logs

{"error":"Review crashed before producing findings. See session logs for details."}

  Session 1 tracking URL: https://claude.ai/code/session_01GExhmd2sRZkNhDRS5QZc5Y
  Session 2 tracking URL: https://claude.ai/code/session_01DwYaG4K3fbVBHYaxP4H7W3

Steps to Reproduce

  1. Open a PR with a substantial number of changed files (Baustein 4.1–5.4,

multi-file React/Node.js codebase)

  1. Run /ultrareview <PR-number> in Claude Code
  2. Wait for the remote agent to finish
  3. Receive task notification with the crash error instead of findings
  4. Repeat – crash occurs consistently on the same PR

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.119 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

The PR contained changes across approximately 10+ files spanning React frontend
(JSX) and a Node.js/Express auth server. The review process appeared to begin
(candidates were identified and verified according to partial output visible
before the crash), but crashed before delivering results.

Two of my three free Ultrareview credits were consumed without producing output.
I would appreciate either a credit refund or confirmation that this is a known
limitation for PRs of this size, along with guidance on the recommended PR size
limit for Ultrareview.

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗