What problem does it solve?
It prevents repeated fix-deploy cycles by forcing a disciplined, hypothesis-driven debugging workflow that verifies changes against the original symptom.
Core Features & Use Cases
- Capture symptoms precisely: Collect the exact URL/endpoint, expected vs observed behavior, a minimal repro recipe, and surface evidence (logs/console/headers) before reasoning.
- Read the architecture first: Inspect the specific files that handle each layer involved in the failing operation rather than guessing.
- Evidence-first hypothesis ladder: Generate 3–5 candidate causes with confirm/refute tests, gather evidence in order, then implement only after confirmation.
- Verify against original evidence: Re-run the same evidence command to prove the fix resolves the symptom and doesn’t introduce a new failure mode.
Quick Start
Invoke /debug and provide a symptom summary including the exact URL or command, expected vs observed results, and the minimal repro steps.