GitHub Agentic Autofix Now Bulk-Fixes Up to 25 Code Quality Findings at Once

What shipped

GitHub shipped agentic autofix for GitHub Code Quality on September 9: select up to 25 findings on a page, assign the whole batch to Copilot in one action, and get a branch with self-validated fixes and a pull request for review.

What changed

The concrete delta is throughput and workflow. The old flow was one finding, one "Generate fix" click; a 200-finding backlog was 200 interactions. The new flow is a page-level batch of 25 — a 200-finding backlog is 8 assignments, and Copilot works them agentically, runs its own validation, then opens PRs. "Assign to Copilot" fully replaces "Generate fix," so there's one code path whether you pick 1 finding or 25.

Three gates matter as much as the feature: bulk remediation requires GitHub Code Quality enabled on Team or Enterprise Cloud (including the data-residency SKU — Hobby and Free repos get nothing); every fix consumes AI credits with no published per-fix rate, so the cost of burning down a backlog is unknowable until you try it; and it inherits your existing enterprise Code Quality policy, so there's no separate approval surface — if the org allows Code Quality, bulk agent fixes are already on.

Why a builder cares

The trust question isn't the fixes — it's the default. A feature that's on wherever Code Quality is enabled means an org that enabled Code Quality months ago just silently gained an agent that opens PRs on its own branches and spends credits doing it. Nobody flipped a switch for that. If your AI credit budget and PR review queue are shared infrastructure, worth a policy check before someone on the team "just clears the backlog" and you find out via the usage line. The review gate is still yours — Copilot validates its own changes, but every batch lands as a PR — so the control exists; the spend just isn't pre-authorized anywhere visible.