draft-bug

You can create a bug report with root cause analysis by gathering code evidence, testing multiple root cause hypotheses in parallel using subagents, and validating them with scratch regression tests to ensure reproducibility.

Evidence-based bug reporting is the process of collecting and analyzing code and log evidence to establish a problem, investigating multiple root causes in parallel with subagents, and validating hypotheses with scratch regression tests. It also includes a blind-spot scan to identify related issues and risks.

Yes, codebase access and the ability to run tests are required for issue tracking and bug reporting. This environment allows the system to execute scratch regression tests to validate root cause hypotheses and generate a comprehensive bug write-up with reproducible steps.

The best way to investigate multiple root causes for a code defect is using parallel subagent investigation. This method tests multiple root cause hypotheses simultaneously, validates them with scratch regression tests, and performs a blind-spot scan to identify related issues and risks.

Regression tests validate a root cause hypothesis during code analysis by serving as scratch tests that reproduce the defect. Running these tests against the codebase confirms or rejects the investigated root causes, ensuring the final bug report contains verified, reproducible regression tests.

You should perform a blind-spot scan during root cause analysis after gathering initial evidence and testing hypotheses. This scan identifies related issues and risks surrounding the primary defect, ensuring the final bug report provides comprehensive coverage of potential regressions.