How to know which tests a PR will break — before CI runs

Short summary

Tests break in CI because reviewers can't see what test code depends on a PR's changes. Static analysis can identify these dependencies (selectors, routes, text) without running the suite, surfacing risks at review time when fixes are cheapest. The author built Testward, a GitHub App that scans test files and uses LLM to predict likely test failures.

• •Static analysis can predict test breakage at PR review time without executing the full suite
• •Key dependencies between app and tests are textual (selectors, routes, labels, IDs) and can be matched programmatically
• •Author's GitHub App (Testward) scans test files and uses LLM confirmation to identify likely spec failures

Generated with AI, which can make mistakes.