G-06Reviewer Agent

A separate agent reviews the diff with fresh context, catching what the author’s context normalized.

Step-by-step walkthrough4 steps

The author’s context has normalized its own mistakes — hours of it.
The reviewer sees only the diff and the repo. No sunk reasoning to defend.
Findings — the bugs the author literally could not see.
Fixes flow back; review repeats until clean. Same model, different context, different blind spots.

Key insight

Author blindness is a context problem, so the fix is a context boundary — the reviewer’s power is precisely what it hasn’t seen.

Failure mode

Sharing the author’s conversation with the reviewer “for context”: you’ve rebuilt the same blind spots and doubled the cost.

Elsewhere in the atlas