What problem does it solve? Research data often looks wrong, noisy, incomplete, or non-convergent, and it is hard to tell whether the cause is a measurement artifact, a preprocessing mistake, a model formulation error, or a solver issue. This Skill provides a structured troubleshooting protocol that isolates the failing layer and turns vague suspicions into falsifiable hypotheses with minimal diagnostic tests. ## Core Features & Use Cases - Data-type branching: Applies distinct checklists for raw sensor/field/SCADA data, simulation and algorithmic outputs (including MCMC chains and power-flow results), and hardware/PLC/communication logs. - Layered failure tree: Separates measurement, preprocessing, model/formulation, solver/algorithm, and reporting layers so symptoms are not confused with root causes. - Minimal experiment design: Converts each suspected cause into one testable hypothesis with an expected result and pass/fail criterion, while enforcing integrity rules against unjustified deletion, smoothing, or imputation. - Use Case: A graduate student whose power-flow simulation fails to converge provides the solver settings, initial conditions, and a baseline case; the Skill identifies likely causes such as tolerance settings or unit mismatches and proposes the smallest controlled test to confirm each one. ## Quick Start Ask the assistant to troubleshoot your research data by describing the data source, the observed issue, the preprocessing steps already done, and your current hypothesis, then request likely causes and minimal diagnostic tests.
