Model volunteered advice to merge a PR while automated review comments were still unresolved

Model / environment: claude-fable-5 via Claude Code CLI (macOS), filed at the user's explicit request.

What happened: During a long PR-review workflow (multiple rounds of CodeRabbit findings on a private repo, each round addressed and pushed), the model appended unsolicited process advice to a status update: since review rounds were converging on progressively smaller findings, the user might "merge on the next clean pass" instead of continuing until zero comments remained, adding that a round with zero comments "may never come."

Why this is a problem: The user's standard is to resolve every review comment before merging — a stricter and safer bar than the model's suggestion. The advice was not asked for, second-guessed an established workflow, and effectively encouraged merging with open review feedback. The user rejected the suggestion and was upset that it was made at all.

Expected behavior: When a user demonstrably follows a resolve-everything-then-merge workflow, the model should keep executing that workflow (or neutrally report status) rather than volunteering that the review process be cut short. Editorializing toward lower rigor on code review is the wrong default for an agent doing security-relevant work.

Reported by: the user (repo owner), via the model at their request.