ultrareview: cloud review crashes before producing findings (2 consecutive runs, same error)

Status Closed — duplicate
Reported on v2.1.198
Maintainer reply None cached
Activity 1 comment · opened Aug 14, 2026 · closed Aug 15, 2026

Summary

/code-review ultra (ultrareview) crashed in the cloud before producing findings on two consecutive runs against the same branch, with the identical error both times. Both runs consumed a free ultrareview credit.

Error reported back to the CLI:

Cloud review failed: orchestrator reported an error: Review crashed before producing findings. See session logs for details.

Environment

  • Claude Code version: 2.1.198
  • Platform: WSL2 (Ubuntu, kernel 6.6.114.1-microsoft-standard-WSL2)
  • Launch mode: /code-review ultra with no arguments, local branch bundle (no PR)

Steps to reproduce

  1. Local git repo, feature branch checked out (branch ~7 commits ahead of main).
  2. Run /code-review ultra (no arguments).
  3. Launch succeeds: "Ultrareview launched for codex/v0.9.2-exhibit-bundle → main (~5-10 min, runs in the cloud)."
  4. After several minutes the task fails with the orchestrator error above. No findings are produced.

Retried once — same crash, same message.

Failed sessions (for log lookup)

Diff scope

  • 58 files changed, 7,092 insertions(+), 693 deletions(-) — well under the documented 500-file / 12,000-line limits
  • Mix of TypeScript source, Playwright specs, and one large JSON Schema file (~1,377 lines). Private repo; no submodules. Large binary files exist in the repo but are untracked and not part of the diff.

Possibly related observation

Before these runs, the no-arg launch was rejected with "Diff is too large" because the base resolved to a stale origin/main rather than local main. Passing /code-review ultra main explicitly produced the byte-identical "too large" message (base argument appeared to still resolve against the remote). Pushing main to origin got the scope check to pass, after which the two crashed runs above occurred. Mentioning in case base/bundle resolution is also involved in the crash.

Impact

Two of three free ultrareview runs consumed by back-to-back infrastructure failures with no findings delivered.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗