What problem does it solve?
Diagnosis helps you stop guessing during an incident by classifying an unknown software failure into the right problem class, collecting the minimum evidence needed to support that classification, and selecting a fitting diagnostic technique before any root-cause debugging begins.
Core Features & Use Cases
- Symptom classification taxonomy (9 classes): Logic Error, Runtime Crash, Data Integrity, Timing / Race, Performance, Configuration, Security, Integration, and Tooling / Build / Script-path.
- Evidence-first diagnostic protocol: Defines universal evidence to collect (exact error/wrong output, reproduction steps, last-known-good state, environment facts) and class-specific evidence checklists.
- Technique-selection matrix: Chooses the cheapest technique compatible with the class (e.g., stack-trace reading for crashes, differential comparison for configuration, profiling for performance, boundary probing for integrations).
- Confidence ladder and reclassification rules: Keeps confidence monotonic with explicit “you can say / you cannot say” boundaries and escalation triggers when stuck.
- Cross-domain patterns: Handles compound scenarios like Cascade, Coincidence, and Environment Ghosts to prevent misleading conclusions.
Quick Start
Use the diagnosis skill to classify an error from its symptoms, collect the required evidence, and route to the most efficient diagnostic approach while tracking confidence and escalation criteria.
