source-control:babysit-prs: advisory-P2-under-approval routing has no ignored-vs-material policy (babysit_feedback.py)

[kyle-sexton]

Lane: babysit. Triggering PR: #567 (fix/499-babysit-snapshot-approve-nits-classification).

Worker's live re-review of the pushed fix (c86fc59) surfaced a design question the code doesn't currently answer: when a bot review is an overall Approve carrying a real P2 finding, should that P2 be routed to material (surfaced to the operator) or silently dropped into the ignored bucket? Currently classify_pr's feedback dict never exposes the ignored bucket at all, so there's no code-backed policy either way — changing it is an author-owned policy call, not a mechanical fix.

Category: behavioral. Needs a human decision on the intended severity-vs-approval-state routing policy before a fix can be written.

[kyle-sexton]

This was generated by AI during triage.

Umbrella-fold → #534 (babysit merge-readiness classifier). "Material vs ignored routing for P2-under-Approve" is finding-state classification — squarely #534's stated pattern ("authorship/finding/approval detection is hand-rolled at each call site; no shared classifier"). Not routed standalone; blocked by #534 recorded natively. The policy call itself (which bucket a P2-under-Approve lands in) still needs an operator ruling — that lives inside #534's shared-classifier design, not as a separate needs-human item.