anthropics / anthropics/claude-code

[Bug] Code review stalls or fails on medium complexity tasks

Open
#92,889 0 comments 0 reactions 0 assignees View on GitHub
area:model bug needs-repro platform:macos
Dominant language
Python
Stars
145k
Forks
23.1k
PR merge metrics
PR metrics pending

Description

**Bug Description**
stalled or failed code review on medium effort with Fable 5.1

**Environment Info**
- Platform: darwin
- Terminal: ghostty
- Version: 2.1.263
- Feedback ID: 20cd3b53-e8ae-4cf6-8e6b-6582af814f92

```
/code-review medium
⎿ Running in the background as @code-review

⏺ Agent "/code-review medium" finished · 3m 47s

The code review is running in the background. So far it has confirmed that formatting and the build are clean, and its finder agents are still analyzing the branch. Findings will be reported once they land and are verified. Nothing to act on yet.

❯ status?

Read 1 file

The review is still in progress, and it has been slow. Here is where it stands:

- Setup done. The review coordinator confirmed formatting and the build are clean on this branch.
- Eight finder agents still running. Three correctness angles plus reuse, simplify, efficiency, altitude, and conventions were launched about 30 minutes ago and none has reported back yet.
- No findings yet. Verification and the final report only start after the finders return.

Thirty minutes without any finder finishing is longer than I would expect for a repo this size, so they may be stalled. I can wait for their notifications, or stop the run and relaunch the review if you would rather not wait.

※ recap: The code review of the branch is still running with no findings yet, since its eight finder agents have not reported after 30 minutes. Next: wait for their results or relaunch the review if they appear stalled. (disable recaps in /config)
```

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing the reported `/code-review medium` run on version 2.1.263, using the feedback ID and environment details from the issue if project tooling supports that. The issue names no files or tests, so first trace the code-review command path and the background finder-agent execution. Done means medium reviews either complete reliably or fail with a clear timeout/error instead of stalling silently.

Written by the indexing model from the issue text.

Assessment

Domain
ai-infra-agents, cli
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.