## What happened `/code-review max` ran as a background forked agent for ~65 minutes, using ~432k tokens and 95 tool calls, without returning any result. It used up almost the entire session limit. - No progress indicator or ETA was shown while it ran. - No warning about expected cost/duration for `max` before starting. - When stopped, no partial findings were surfaced — confirmed findings existed only inside the raw task output file. ## Expected - A cost/duration warning (or cap) for `max` ef

/code-review max ran ~65 min / ~432k tokens with no progress and no result, consuming the session limit

Status Open

Reported on v2.1.281

Maintainer reply None cached

Activity 0 comments · opened Sep 24, 2026

What happened

/code-review max ran as a background forked agent for ~65 minutes, using ~432k tokens and 95 tool calls, without returning any result. It used up almost the entire session limit.

No progress indicator or ETA was shown while it ran.
No warning about expected cost/duration for max before starting.
When stopped, no partial findings were surfaced — confirmed findings existed only inside the raw task output file.

Expected

A cost/duration warning (or cap) for max effort, especially near the session limit.
Visible progress while the background review runs.
Partial, already-verified findings returned when the run is stopped or the limit is near.

Repro

On a branch with several infra commits (~multi-file TypeScript/Pulumi + Java diff), run /code-review max.
Wait. The agent runs for over an hour with no feedback and consumes most of the session budget.

Environment

Claude Code 2.1.281, macOS
Date: 2026-09-24

Refund

This run delivered no usable result and consumed almost the entire session. I'm expecting a refund / usage credit for the tokens spent on it (~432k tokens).

View original on GitHub ↗