GitHub has updated Copilot code review to resolve its own comments when a later commit addresses the underlying feedback, leaving outstanding comments open for developers who still need to act. It also generates change-specific commit messages when users apply Copilot autofix suggestions, replacing the standard autofilled message.
Copilot can now validate reviewed code with the full set of shell tools from the Copilot SDK, operating behind the Copilot agent firewall. GitHub says this can include running builds and tests, executing targeted scripts, and retrieving information through available tools and APIs. These analysis changes do not alter how developers request or receive reviews.
GitHub reported that its experiments with the expanded tool access produced more positive developer feedback, more high-severity findings and fewer minor nits. These outcomes are GitHub’s experimental results rather than independently validated measurements.
The Lite effort level now assigns reviews to an ensemble of agents, combines their different findings and presents the result as one review. In GitHub’s experiments, the approach increased the average number of addressed comments per review by 47% for high-severity findings, 31% for medium-severity findings and 11% for low-severity findings, while cutting review cost by about 8%.
Source: GitHub Changelog.
Definition. GitHub Copilot code review is a review workflow that analyzes code, provides feedback and can now close its own comments after later commits address the underlying issue.
| GitHub experimental metric | Reported result |
|---|---|
| Addressed high-severity comments per review | Increased 47% |
| Addressed medium-severity comments per review | Increased 31% |
| Addressed low-severity comments per review | Increased 11% |
| Review cost | Decreased about 8% |
Key takeaways
- Copilot now resolves its own review comments when later commits address the underlying feedback.
- Outstanding comments remain open when developers still need to act.
- Applied autofix suggestions receive change-specific commit messages instead of the standard autofilled message.
- Copilot can use the Copilot SDK shell tools behind the Copilot agent firewall to run builds, tests and targeted scripts.
- Lite reviews now combine findings from an ensemble of agents into one review.
- GitHub characterized the performance figures as experimental results, not independently validated measurements.
FAQ
When does Copilot automatically resolve a review comment?
Copilot resolves its own comment when a later commit addresses the underlying feedback; comments requiring further developer action remain open.
What changed about Copilot autofix commit messages?
Copilot now creates a change-specific commit message when a user applies an autofix suggestion instead of using the standard autofilled message.
What tools can Copilot use during code review?
It can validate reviewed code with the Copilot SDK shell tools behind the Copilot agent firewall, including builds, tests, targeted scripts and retrieval through available tools and APIs.
How do Lite reviews use multiple agents?
The Lite effort level assigns a review to an ensemble of agents, combines their different findings and presents them as one review.
What results did GitHub report for multi-agent Lite reviews?
GitHub reported increases in addressed comments of 47% for high-severity findings, 31% for medium-severity findings and 11% for low-severity findings, alongside about 8% lower review cost.
Are GitHub's reported improvements independently validated?
No. The post identifies them as GitHub's experimental results rather than independently validated measurements.
