Configurable path excludes for PR-comment Bugbot (size gate + review context)

Feature request for product/service

BugBot

Describe the request

For GitHub PR-comment Bugbot (bugbot run / cursor review), large required files can block or dilute review even when almost none of the real findings live in them.

Examples of files that often must ship in a PR but add little review signal:

  • ORM / DB migration metadata (e.g. Drizzle *_snapshot.json, _journal.json)
  • lockfiles and dependency graphs (when not already filtered)
  • codegen output (OpenAPI/GraphQL/protobuf clients, Prisma artifacts)
  • large snapshot / fixture JSON

Today the size gate can refuse the whole PR (BUGBOT_TOO_LARGE_TO_REVIEW), and when review does run those blobs still burn tokens. Repo rules / BUGBOT.md path scopes steer what to comment on, not which files load or count toward the gate. .cursorignore is IDE-only. Stacking onto a non-default base also doesn’t help if the gate is vs the default branch.

Ask: repo-level path excludes (e.g. .cursor/bugbotignore or equivalent config) applied before the size gate and context load. Optionally size against the PR base after excludes, and document it for monorepos.

Concrete case: a ~134k-line PR where ~130k lines are Drizzle migration metadata; the real code delta is a few thousand lines and would review fine without those files in the gate/context.

Hi @David_Stredansky — thanks for the detailed write-up.

The expected behavior is that marking files as linguist-generated=true in .gitattributes is enough to exclude them from Bugbot review. However, there’s a known issue where this isn’t working as intended, and I’ve attached your report to it so we can update you when it’s fixed!