Key GitHub PR Review Bot Features Worth Leveraging

Batched/Pending Reviews

Create a review, attach all inline comments, then submit once. This fires a single notification instead of one per comment and keeps the review atomic. Use POST /repos/…/pulls/…/reviews with event: REQUEST_CHANGES or APPROVE.

Multi-line Suggestion Blocks

Suggestions can span multiple lines. Combine with the start_line / line parameters when creating review comments. Critical for refactor suggestions.

Review Thread Resolution

Bots can programmatically resolve/unresolve comment threads via GraphQL (resolveReviewThread). Useful for auto-resolving threads when a follow-up commit fixes the flagged issue.

Review State Transitions

REQUEST_CHANGES blocks merge (if branch protection requires it). A bot can APPROVE or dismiss its own previous reviews when conditions change — giving you a real merge gate, not just informational comments.

PR Description Mutation

Bots can rewrite the PR body with structured summaries, checklists, or risk assessments on each push. Pairs well with a summary section the bot owns (delimit it with HTML comments so you can find/replace it).

File-level Comments

Comments without a line reference attach to the file card rather than a specific line. Useful for "this entire file needs X treatment" observations.

Draft PR Detection

Check draft: true and skip expensive reviews; re-trigger on the ready_for_review event. Saves API quota and avoids noise.

Highest-Leverage Combination

Batched review + suggestion blocks + check run annotations — reviews own the conversational feedback, annotations own the machine-generated lint/security findings.