What is Cross-PR Analysis?
Last updated
Cross-PR analysis is the practice of evaluating how multiple open or recent pull requests interact with each other — detecting conflicting changes, shared dependency modifications, and cumulative architectural drift that only becomes visible when you look across PRs rather than at each one in isolation.
Why does cross-PR analysis matter for engineering teams?
Individual PRs look safe in isolation. But when three PRs each modify a shared interface in compatible but different ways, the combined result can break it. Cross-PR analysis looks for these interaction effects, which a review of one PR at a time cannot see.
How does Argus handle cross-PR analysis?
Argus checks only PRs linked in the PR description, as GitHub PR URLs or owner/repo#N (up to 5 by default). It does not scan other open PRs on its own. After the main review, one LLM call reads the linked PRs' diffs and their earlier Argus findings and looks for risks that appear only when the changes combine, across 9 categories such as schema races, serialization contracts, type drift, and deploy ordering. Results are added to Argus's review as an informational section. When 2 or more linked PRs reference the same issue, Argus also judges that issue's acceptance criteria against the combined change.