Can I run multiple trials for a single scenario?

Yes, you can use the --trials N argument to run a scenario multiple times, which is useful for gathering statistical averages and ensuring consistency.

What is the purpose of Nabledge-Test?

It is a benchmarking framework designed to measure the performance and accuracy of Nablarch-related skills in Claude Code, ensuring they meet specific domain requirements.

Which metrics are tracked during execution?

The framework tracks total execution time, token usage (input/output estimates), the number of tool calls (Read/Bash/Grep/Write), and keyword detection rates.

What does Baseline Mode do?

Baseline Mode runs all scenarios, saves the results as a reference point, and generates a comparison report to identify improvements or regressions in performance.

Why does it run scenarios in isolated sub-agents?

Isolation ensures that the context from one test doesn't influence the next, providing an unbiased measurement of how the AI performs on a single specific task.

Nabledge Skill Benchmark Framework

Name: Nabledge Skill Benchmark Framework
Author: nablarch

bynablarch

0•

Seguridad y Pruebas

Benchmarks and validates Nablarch-specific AI skills using isolated sub-agent execution to ensure high-precision performance and regression detection.

Nabledge-Test is a specialized benchmarking framework designed to evaluate the accuracy and efficiency of Nablarch development skills within Claude Code. By executing test scenarios in isolated sub-agent contexts, it eliminates cross-scenario bias and provides a rigorous measurement environment. The tool tracks critical metrics including keyword detection rates, token consumption, and execution time, allowing developers to generate baseline reports and verify improvements or regressions in their domain-specific AI workflows.

Características Principales

01Automated keyword and component detection for QA and Code Analysis

02Isolated sub-agent execution to prevent context contamination

03Support for multi-trial statistical analysis to ensure result stability

04Baseline comparison mode for regression testing and performance tracking

050 GitHub stars

06Granular metrics reporting including token usage and tool call counts

Casos de Uso

01Performance profiling of complex code analysis workflows

02Validating the accuracy of AI-generated Nablarch architecture diagrams

03Regression testing after updating Nablarch-specific knowledge bases

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nablarch/nabledge-dev nabledge-test

For use in Claude.ai and ChatGPT

Download Skill