claude-code - 💡(How to fix) Fix /ultrareview returns empty findings on large diff (105 files / +19k LoC) — two consecutive runs [1 comments, 2 participants]

Summary

/ultrareview returned an empty findings array ([]) on two consecutive runs against the same branch, against a sizable diff (~105 files, +18,967 / -1,044). Both runs reported status: completed with zero findings — no notes, no rationale, no partial coverage indicator.

Sessions

• session_01ADykSHyLkvNizg3Co1vTpf
• session_01AswYbXxdnjCLMY2jNDfr7Y

Reproduction

1. Checkout a feature branch with a large diff vs main (~100+ files, ~20k LoC, mix of infra/service/test/docs).
2. Run /ultrareview.
3. Wait for the task-notification.
4. Observe findings: [].
5. Repeat — same result.

Expected behavior

One of:

• Surface "no critical findings" with a one-line rationale, or
• Surface a "scope exceeded analyzer ceiling — partial coverage" notice with what was/wasn't covered, or
• Return at least style/design observations on a diff this large.

A bare [] with no metadata gives the user no signal on whether the analyzer ran end-to-end, hit a ceiling, or genuinely found nothing.

Impact

• Two of three monthly free runs consumed without actionable output.
• User has to manually decide whether to retry or skip the third run.

Asks

1. Investigate whether the analyzer silently truncated/aborted on this scope (logs from the two session IDs above).
2. If a known scope-ceiling exists, surface it explicitly in the result instead of [].
3. Consider refunding the two consumed free runs once cause is confirmed.

Environment

• Claude Code CLI, model Opus 4.7 (1M context)
• macOS Darwin 25.3.0
• Diff scope: 105 files changed, +18,967 / −1,044