fix(ci): exempt draft PRs from the opencode-review required-check poll

[seonghobae]

Summary

#1443 fixed the required opencode-review-target check hanging forever on a draft PR. It was closed unmerged after #1546's receipt-gate redesign landed, on the premise that the redesign already handles drafts via pr_review_merge_scheduler.py's draft path and opencode_review_receipt_gate.py's is_draft parameter.

That premise is only half right. Reproduced against current main (5686de41):

• evaluate_receipts's is_draft only narrows what counts as a valid receipt (if is_draft and state == "APPROVED": return False, "draft must never receive bot APPROVE") — it never exempts a draft PR from needing a receipt at all.
• pr_review_merge_scheduler.py's own draft path (inspect_pr's if pr.get("isDraft")) skips dispatching a review entirely for an ordinary draft with no @opencode-agent mention (active_draft_review_request is documented as "the sole automatic gate for draft review dispatch").
• Net effect: for an ordinary draft PR, nothing ever posts a verdict, and the Fail closed without a current-head OpenCode verdict step's while :; do ... sleep 30; done loop has no draft check at all — it polls until the job's own ~360-minute runtime ceiling kills it.

#1443's underlying bug still reproduces on current main.

Fix

Per that closure's own guidance ("any residual draft-queue issue must be reproduced against current main and fixed in the current receipt/scheduler boundary rather than reviving this stale branch"), this is a fresh, minimal fix on a new branch off current main — not a revival of #1443's branch:

• The Fail closed step now also reads PR_DRAFT: ${{ github.event.pull_request.draft }}, matching the sibling dispatch step's existing sourcing convention exactly (deliberately not the live gh api refetch #1443's branch had introduced).
• It exits early on a draft PR, mirroring its pre-existing closed exit.

The receipt-gate/scheduler architecture itself is otherwise untouched — this is scoped to the one missing exemption.

Test plan

• Added test_fail_closed_step_exempts_a_draft_pr_before_polling, test_fail_closed_step_closed_still_takes_precedence_over_draft, test_fail_closed_step_still_polls_for_a_non_draft_pr — each executes the actual production step body via a fake gh that fails loudly if invoked, proving the draft/closed exits short-circuit before any Reviews API call (and that a non-draft PR still reaches it).
• tests/test_opencode_required_verdict_regression.py — 25/25 passed.
• Full suite (coverage run -m pytest tests) — 2249 passed, 1 skipped, 21 subtests, matching vanilla main's 2246 passed + 3 new tests.
• Confirmed via git stash that the suite's one pre-existing coverage shortfall (99% vs 100%, in pr_review_fix_scheduler.py/pr_review_merge_scheduler.py — files this PR does not touch) is present on vanilla main too, unrelated to this change.
• ruff check on the modified test file — clean. interrogate on the modified test file — 100%.
• YAML parses; both run: script bodies pass bash -n.

---

Generated by Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

The code fix remains valid and current-main-based, but this PR is Draft and the connected markPullRequestReadyForReview mutation is still broken by the connector requesting nonexistent GraphQL field Repository.fullDatabaseId. To avoid leaving a queue-amplification fix permanently unreviewable, I am replacing only the PR conversation with a ready successor from the identical unchanged branch/head. No code/history changes and no #1560 checks/reviews will transfer.