systematic-debugging

Test-first debugging enforces writing a failing test that reproduces the bug before attempting any fixes. It reads bugfix items from design docs, confirms the failure by running the test suite, and then proposes verified corrections.

To debug using parallel subagents, fetch the current bugfix item from the design doc, write a failing test, and spawn 2 to 3 subagents concurrently. These subagents explore and propose fixes simultaneously to surface the best approach.

Yes, subagents can reproduce and fix issues described in design documents. After a failing test confirms the bug, parallel subagents are spawned to explore fixes. Upon verification, the design doc and session state are updated with the confirmed fix.

The best way to coordinate multiple debugging approaches is spawning parallel subagents. This workflow runs 2 to 3 subagents concurrently to explore proposed fixes, allowing you to compare approaches and select the verified best solution.

Yes, a design doc is required. The debugging cycle reads the current bugfix item directly from the design doc to understand the issue, and later updates that same design doc with the confirmed fix and documented rationale.

Writing a failing test before a bugfix ensures accurate reproduction of the issue. Running the test suite to confirm the failure prevents false positives, ensuring that any proposed fix from subagents actually resolves the targeted problem.