pr-and-verifylisted

# Open the PR and drive it to proven green This is where a finished loop hands off. A work-item that settled `DONE` or `DONE_WITH_CONCERNS` in its worktree has code that passes its own check *and* already carries a clean review verdict earned inside the loop. This skill opens the PR only once that verdict is on record, then drives CI and the contract's observable done-criteria to green that is **proven end-to-end, not asserted**. It never opens a PR on an unreviewed change, never lands one on a red check, and never lands one on a done-criterion nobody watched go green. Like the loop it follows, it leans on the other better-dev practices rather than re-doing their jobs. ## What it leans on - **The contract** - `.better-dev/bin/bd-mem ledger read <work-item> contract.md` holds the observable done-criteria and the red-capable signal the loop graded against. That is what "verified" means here, not "the unit tests are green." - **The verdict** - `/review` runs *inside the loop*, before DONE, from a fresh context that never sees the report, and records a clean result to the work-item's ledger. This skill reads that record; it never re-runs the review. The findings, every severity of them, were the loop's to clear, not this skill's to fix. - **The fix loop** - review findings and red CI both go back to `/autonomous-loop`, which owns the implement-and-verify loop. This skill decides *when* the change is not yet green; it does not run a second fix loop of its own. - **T