investigate

To find the root cause of a 500 error, use a systematic four-phase workflow that enforces context recovery and root cause confirmation before any fixes are applied. This prevents recurring issues by validating scope boundaries and logging telemetry.

The best way to debug production incidents is to follow a strict sequence of investigating, analyzing, hypothesizing, and implementing. This systematic approach restores recent decisions and cross-session learnings to prevent whack-a-mole debugging.

To troubleshoot stack traces without losing progress, apply context recovery to restore recent checkpoints and learnings after context window compaction. This ensures your debugging session maintains all accumulated investigation data.

You cannot apply fixes immediately when debugging unexpected behavior due to the Iron Law, which enforces that no fixes happen without root cause confirmation. Pre-tool hooks validate debug boundaries via freeze checks before any edits are allowed.

This systematic debugging approach works directly for production incidents in web applications and services. It applies to 500 errors, stack traces, unexpected behavior, and regressions, enforcing scope boundary validation to prevent recurring issues.

You need scope boundary validation before editing code to prevent uncontrolled changes during bug investigation. Pre-tool hooks enforce freeze checks before any edits or writes are allowed, ensuring fixes target confirmed root causes safely.