Research Results Auditor

Produce a structured validity verdict for completed experiment results before they become paper claims. This skill asks: is the evidence scientifically defensible as stated?

Use this skill when:

• experiments are complete and results are about to be written into the paper as claims
• a result looks convincing but the connection between the measurement and the stated conclusion feels stretched
• a reviewer is likely to ask "what else could explain this?"
• an ablation result is used to attribute a performance gain to a specific component
• class imbalance, calibration, or threshold choice may be driving the metric
• the gap between what was measured and what is claimed in the paper needs to be checked before submission

Do not use this skill when experiments are still running or results are not yet final — use result-diagnosis for in-progress diagnosis and next-action decisions. Do not use this skill for test selection or variance reporting — use statistical-analysis-planner.

Pair this skill with: