mma-debug

Cross-file debugging automates reading, reproducing, and tracing failures across multiple files using a worker-based hypothesis workflow to isolate root causes. It delegates investigative work to an isolated worker while the main agent focuses on decisions.

Tracing broken builds across components is handled by a hypothesis-driven workflow that delegates investigative tasks to an isolated worker. This systematic approach keeps the main hypothesis intact while reproducing symptoms and reading affected files.

Hypothesis-driven debugging is needed when failures span more than one file or component, requiring cross-file reasoning and reproduction. It coordinates reading files and tracing across modules to isolate the root cause of unexpected runtime behavior.

To start cross-file debugging, provide a failing test, reproduction steps, and affected files to the worker. The worker runs in an isolated context to investigate symptoms and trace the root cause without polluting results.

Isolated worker context ensures reproducibility and avoids polluted results during debugging. Each investigative task runs independently, allowing the main agent to maintain a clean hypothesis while tracing failures across multiple modules.

Automated cross-file debugging requires clear reproduction steps and affected files to function effectively. Without accurate initial inputs for the failing test, the isolated worker may struggle to trace the root cause across modules.