skills.rest
MichaelsEngineering
MichaelsEngineeringCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

gpd-compare-results

Compares internal results, baselines, or methods and emits machine-readable verdict artifacts.

Updated Mar 18, 2026
Search Tags:#validation#comparison#benchmark#baseline#verdict#gpd#physics research
One-click install
npx skills add https://github.com/MichaelsEngineering/sentinel-multiscale-field-dynamics --skill gpd-compare-results-michaelsengineering
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: gpd-compare-results
Source: https://github.com/MichaelsEngineering/sentinel-multiscale-field-dynamics/tree/main/.agents/skills/gpd-compare-results
Command: npx skills add https://github.com/MichaelsEngineering/sentinel-multiscale-field-dynamics --skill gpd-compare-results-michaelsengineering

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? It replaces vague prose comparisons like "the two plots look similar" with explicit, machine-readable comparison artifacts that record metrics, thresholds, verdicts, and follow-up actions for load-bearing scientific results. ## Core Features & Use Cases - Decisive Comparisons: Compares analytics vs numerics, method A vs method B, benchmark vs produced quantity, or baseline vs modified run with explicit metrics and thresholds. - Verdict Classification: Classifies each comparison as pass, tension, fail, or inconclusive and routes follow-up debugging or verification actions. - Contract-Backed Linkage: Resolves subject_id, subject_kind, and reference_id when the comparison target is contract-backed, and writes artifacts to .gpd/comparisons/ using the internal-comparison template. - Use Case: After running a turbulence baseline rollout, compare the produced energy spectrum against the benchmark dataset, emit a comparison_verdicts artifact, and route a tension verdict to a bounded debugging step. ## Quick Start Ask the AI to compare the current run's output against the baseline benchmark and emit a decisive comparison verdict artifact.

Frequently Asked Questions about gpd-compare-results

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I compare two simulation results with explicit pass/fail verdicts?▼

Run the compare-results workflow with the two sources identified, a shared metric such as relative error or chi-squared, and a threshold. It computes the metric, classifies the verdict as pass, tension, fail, or inconclusive, and writes a comparison artifact.

What is a machine-readable comparison artifact in GPD?▼

It is a markdown file written to .gpd/comparisons/ using the internal-comparison template, with frontmatter containing comparison_verdicts, comparison_sources, and subject_id or reference_id linkage when the target is contract-backed.

Can I compare a baseline run against a modified run?▼

Yes, baseline-vs-modified is a supported comparison mode alongside analytics-vs-numerics and cross-method checks. Both sides must represent the same observable with matching conventions and normalization before a verdict is computed.

Why does the comparison workflow stop at initialization?▼

The workflow halts if the GPD runtime CLI init progress command fails, displaying the error instead of proceeding. Verify the pinned runtime invocation and the ./.codex config directory exist before rerunning.

What happens when a comparison verdict is inconclusive?▼

An inconclusive verdict triggers explicit identification of the missing data or missing normalization needed to make the comparison decisive, rather than leaving the result implicit in prose.

Related Skills

Explore More Education & Research Skills→
standardbeaglestandardbeagle

audit-benchmark-evidence

Audit benchmark validity, reproducibility, and evidence claims from raw artifacts.

Official
Advanced
LightconeResearchLightconeResearch
18

figure-comparison

Compares paper reference figures, tables, and values with reproduced results.

Official
Advanced
DozyboyDozyboy

hs:critique

Fans independent review lenses at an artifact and consolidates findings into one ranked verdict.

Community
Advanced
HohlasHohlas
3

create-eda-report

Run a statistics pipeline and EDA notebook to generate JSON/CSV reports and plots.

Community
Intermediate