[BUG] `/ultrareview` fails at Verify stage repeatedly, consumes free quota, drops all findings

Resolved 💬 3 comments Opened May 3, 2026 by alirexha Closed May 13, 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?

Two consecutive /ultrareview runs on the same branch failed at the Verify stage. Both runs:

  • Successfully completed Setup and Find (returning 6 and 9 candidates respectively)
  • Crashed/failed at Verify
  • Never ran Dedupe
  • Returned an empty [] array of findings to the local Claude Code session
  • Counted against my free quota (now 2 of 3 used)

Result: two of three free ultrareviews consumed with zero findings delivered, despite the reviewer clearly having identified candidate issues.

Environment

  • Platform: CachyOS (Linux 7.0.3-1-cachyos)
  • Claude Code: Research preview (Max plan)
  • Date: 2026-05-03

Run 1

  • Session URL: https://claude.ai/code/session_017Vp61ZHEL6CSkfm5DxLYuQ
  • Branch: feat/phase-5-paritymain
  • Scope: 54 files changed, 19,637 insertions(+), 1,405 deletions(-)
  • Status: Setup ✅ → Find ✅ (6 candidates) → Verify ❌ → Dedupe (skipped)
  • Local result: []
  • Quota counter at launch: "Free ultrareview 1 of 3"

Run 2

  • Session URL: https://claude.ai/code/session_01CDCuShCVpC5BF4ttQ1y57c
  • Branch: feat/phase-5-paritymain (same)
  • Scope: same
  • Status: Setup ✅ → Find ✅ (9 candidates) → Verify ❌ → Dedupe (skipped)
  • Local result: []
  • Quota counter at launch: "Free ultrareview 2 of 3"

What Should Happen?

Either:

  1. Failed runs should not count against the free quota, OR
  2. Candidate findings from Find should still be surfaced when Verify fails (with a clear "unverified" tag), OR
  3. The pipeline should not silently emit [] to the user when an upstream stage failed.

Error Messages/Logs

Steps to Reproduce

  • Screenshots of both failed runs in the ultrareview UI showing "Review failed" with candidate lists visible.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.126

Platform

Anthropic API

Operating System

Other Linux

Terminal/Shell

Terminal.app (macOS)

Additional Information

<img width="1920" height="1080" alt="Image" src="https://github.com/user-attachments/assets/fe0c936a-4371-4f48-bc1d-2150d9d04eed" />

<img width="1920" height="1080" alt="Image" src="https://github.com/user-attachments/assets/6e4e538e-0d5b-4596-bc96-8dd2aaed5bfc" />

View original on GitHub ↗

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