principle-explain-the-number
backnotprop1.6k installs
Make measured results stand up to scrutiny.
What it does
Checks measured results against runtime evidence so reported numbers have an identified limit and alternative explanations ruled out.
When to use it
Use before trusting, reporting, or acting on a measured performance number or evaluation result.
How to use it
It changes how the agent validates and reports measured results, requiring runtime evidence, alternative explanations checked, and supporting notes.
What you provide
- An existing project
Latest review
What you need · 1
Requires the benchmark-checklist skill to run the full procedure for performance numbers; evaluation results instead receive the stated checks on trials.
Access · 0
No additional actions listed.