fix: draft-PR guard on the caller

[cansin]

One line, long overdue: the mirror's header documents the draft gate as the caller's job, but the caller never carried it — drafts got blocking self-hosted reviews. Surfaced when bly's first Claude review flagged the identical omission in the caller copied from here.

🤖 Generated with Claude Code

https://claude.ai/code/session_01LjEWDajYkRMzCdEUhw44jc

[github-actions]

Automated review by claude (mode: agentic)

Verdict: PASS — The review found no must-fix issues, describing the change as a correct, minimal draft-gate addition.

No issues found — the change is a correct, minimal draft-gate addition.

[github-actions]

Automated review by claude (mode: agentic)

Verdict: FAIL — The workflow trigger omits ready_for_review, so draft PRs marked ready without a new push never get the required check and can permanently block merge under the documented branch-protection setup.

One finding: the new draft gate skips the job on opened for drafts, but the workflow's trigger types: [opened, synchronize] omits ready_for_review — so a draft PR marked ready without a subsequent push never gets the required claude-review / PR review check, which can permanently block merge under the required-check branch protection described in the file's own header comment.

[github-actions]

Automated review by claude (mode: agentic)

Verdict: PASS — The review found no correctness bugs, security issues, or violated invariants in the draft-PR gating change to the workflow file.

Reviewed the full diff — it's a small, self-contained change to .github/workflows/claude-code-review.yml:

1. Adds reopened and ready_for_review to the pull_request trigger types.
2. Adds if: ${{ github.event.pull_request.draft == false }} to the claude-review job (which calls the reusable _claude-review.yml via uses:).

Logic check:

• draft is a genuine boolean in the pull_request payload, so == false compares correctly (no stringly-typed footgun).
• Skipping a job with if: on a uses:-based (reusable-workflow) job correctly prevents the reusable workflow from being instantiated at all, rather than reporting a skipped/green check — matches the comment's claim that this avoids poisoning a required check with a false pass.
• ready_for_review firing sets draft: false in that event's payload, so adding it to types combined with the new if gate does mint the check exactly once when a draft is published without a new commit — closing the gap described in the comments.
• Draft PRs on opened/synchronize are correctly skipped (no premature check run), and non-draft PRs on all four listed events still run.

No correctness bugs found: the trigger-list change and the job-level draft gate are consistent with each other and with the stated intent, and I don't see an edge case in the four event types where draft would be stale or misreported at fire time.