[BUG] /ultrareview run failed

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?

My /ultrareview run on PR #7 of my private repo crashed tonight before producing any findings.

Expected behavior: The review returns a structured findings report (file:line, severity, recommended fix, etc.), and the billed amount reflects work that produced output.

Actual behavior: After ~several minutes, the task completed with status "completed" but the findings payload was an error object: {"error": "Review crashed before producing findings. See session logs for details."}

A usage charge of $40+ was incurred for the failed run despite no findings being produced.

Context — the PR that was reviewed:

• Repo: cever-alliance/alliance-accommodations (private)
• PR: #7 — "Module 1 / Auth: Google SSO + first-sign-in provisioning + route protection"
• URL: https://github.com/cever-alliance/alliance-accommodations/pull/7
• Base branch: main
• Head branch: claude/great-wozniak-6c7139
• Head commit SHA: 6c9ee2577f91634e21ab44daa96502099e69ed66
• Size: 21 files changed, +2,922 / -3 lines
• Stack: Next.js 16 App Router, TypeScript, Supabase (@supabase/ssr), Vitest

Session identifiers for log lookup:

• Ultrareview task ID: rbv805cug
• Linked session URL printed by the launch command: https://claude.ai/code/session_011uMBsRoCK2D8a1i6CbpuPB

What Should Happen?

The review should return a structured findings report (file:line, severity, recommended fix, etc.), and the billed amount should reflect work that produced output.

Error Messages/Logs

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

Steps to Reproduce

Steps to reproduce:

1. Inside Claude Code (CLI v2.1.143, macOS 26.4.1), within a project that has the gh CLI authenticated to a GitHub account with access to a private repo containing an open PR.
2. At the prompt, type the slash command: /ultrareview 7 (where 7 is the PR number).

Confirm the billing prompt ("This review bills as usage credits.").

1. Wait for the remote agent to complete (~5–10 minutes per the launch message).

Expected behavior: The review returns a structured findings report (file:line, severity, recommended fix, etc.), and the billed amount reflects work that produced output.

Actual behavior: After ~several minutes, the task completed with status "completed" but the findings payload was an error object: {"error": "Review crashed before producing findings. See session logs for details."}
A usage charge of $40+ was incurred for the failed run despite no findings being produced.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

1.9255.0

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

I ran the standard slash command (/ultrareview <PR#>) once against an open PR in a private repo and received the crash on the first attempt. I have not retried because of the billing exposure. If the engineering team needs to reproduce against this specific PR's contents, I'm willing to grant access — please reach out via support and I'll add the necessary GitHub user(s) as collaborators.