GitHub has updated Copilot code review with both workflow and analysis changes. When a later commit addresses a Copilot review comment, Copilot can now resolve that thread during rereview instead of leaving developers to close it manually.
Applying a Copilot suggestion also generates a commit message tailored to the change. On the analysis side, the reviewer can use the broader shell-tool set from the Copilot SDK behind GitHub's agent firewall, including build commands, tests and targeted scripts where available.
The Lite effort level now uses an ensemble of agents rather than a single reviewer. GitHub reports that its experiments increased addressed high-severity comments by 47%, medium-severity comments by 31% and low-severity comments by 11%, while reducing review cost by about 8%. Those figures are first-party experimental results and are not treated as independent comparative evidence.
For AiToolMap, this is a material GitHub Copilot update because it changes both the review engine and the developer workflow after feedback is applied. The existing review should be revalidated for the new multi-agent analysis path, while retaining clear boundaries between GitHub's own experiment results and independent evidence.