GitHub Copilot Code Review Now Auto-Resolves Addressed Comments and Writes Smarter Commit Messages

GitHub Copilot’s code review feature gained automatic comment resolution and several analysis improvements aimed at surfacing higher-quality findings.

Details

• Auto-resolution: Copilot now closes its own review comments automatically once a subsequent commit addresses them, so only genuinely outstanding feedback stays visible instead of requiring manual thread management
• Shell tools integration: Copilot now has access to the full suite of shell tools from the Copilot SDK, letting it validate code through build commands, test runs, and scripts — resulting in more high-severity findings and fewer minor, low-value ones
• Ensemble review at Lite effort: the Lite effort level now runs multiple agents collaborating on a single review; GitHub says this increased the average number of addressed comments per review by 47% for high-severity findings while cutting cost by 8%
• Smart commit messages: applying one of Copilot’s suggested fixes now generates a commit message reflecting the actual change made, replacing generic auto-filled templates

What happened next

The changes push Copilot’s code review toward behaving more like an attentive human reviewer — tracking what’s actually still open, validating claims by running real build and test commands rather than just reading code, and writing commit messages that describe what changed — while the ensemble approach at the cheaper Lite tier suggests GitHub is finding ways to improve review quality without proportionally raising cost.