investigate

Root-cause debugging enforces a four-phase workflow—investigate, analyze, hypothesize, implement—to systematically resolve stack traces and intermittent failures. It collects evidence like log fragments and diffs to validate hypotheses, ensuring durable fixes rather than superficial patches.

Hypothesis-testing debugging is a structured approach that requires forming and validating hypotheses with concrete evidence before applying fixes. By using an iron law guardrail that blocks fixes without root-cause confirmation, it eliminates guesswork and prevents recurring errors.

To start an investigation workflow, invoke the agent to begin the structured debugging process. The workflow gathers data using tools like Bash, Grep, and Read to collect symptoms and traces across modules, guiding you sequentially through the investigation and analysis phases.

Yes, phase-based debugging requires a project-local .tmp workspace to function correctly. This workspace serves as a scratch area where the automation tools can write intermediate findings, collect log fragments, and store trace data during the investigation phases.

The best way to debug intermittent failures is using a structured evidence-gathering workflow that collects symptoms, log fragments, and diffs at each step. This phase-based approach validates hypotheses with concrete data before implementation, ensuring the root cause is confirmed.

Yes, this systematic debugging workflow applies to errors, stack traces, and intermittent failures across multiple modules. It uses Bash, Grep, and Glob tools to collect data project-wide, forming and verifying hypotheses to reach durable fixes regardless of the module boundaries.