benchmark-suite

Coordinate multi-lane benchmark lifecycles across providers and models.

Updated Mar 26, 2026
One-click install
npx skills add https://github.com/assafYavnai/ADF-V1 --skill benchmark-suite
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: benchmark-suite
Source: https://github.com/assafYavnai/ADF-V1/tree/main/skills/benchmark-suite
Command: npx skills add https://github.com/assafYavnai/ADF-V1 --skill benchmark-suite

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Coordinating and supervising multi-lane benchmarks across a provider/model/runtime matrix with isolation, KPI capture, and Brain persistence.

Core Features & Use Cases

  • Orchestrates end-to-end benchmark lifecycles across multiple lanes (implement-plan → machine verification → review-cycle) with per-lane isolation and run control.
  • Captures KPIs, artifacts, and summaries for each lane and the whole suite.
  • Use Case: Run a 3-provider, 4-model benchmark matrix to compare performance across configurations.

Quick Start

Start a new benchmark suite by providing a valid config JSON; the system will spin up lanes, run them, and generate a final summary.

Frequently Asked Questions about benchmark-suite

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

FAQPage Schema
How do I orchestrate multi-lane benchmarks across different AI providers and models?

Multi-lane benchmark orchestration coordinates an isolated lane matrix across providers and models through a governed cycle of implement-plan, machine verification, and review-cycle with KPI capture. You provide a config JSON to spin up lanes, run them, and generate a final summary.

What is the best way to capture KPIs for a multi-provider AI benchmark suite?

KPI capture for a multi-provider benchmark suite records metrics, artifacts, and summaries for each lane and the entire run. The workflow supervises the isolated lanes through a governed cycle and persists the captured KPI data into the Brain.

Can I run a 3-provider, 4-model benchmark matrix with per-lane isolation?

Yes, you can run a 3-provider, 4-model benchmark matrix with per-lane isolation. The suite orchestrates the end-to-end benchmark lifecycle by validating inputs, loading instructions into context, and enforcing safety and execution rules across the isolated lanes.

What do I need to start a benchmark suite and run a governed review cycle?

To start a benchmark suite and run a governed review cycle, you need a valid config JSON. The system uses this configuration to spin up isolated lanes, execute the implement-plan to machine verification workflow, and enforce required execution rules.

Why does the benchmark workflow require machine verification before the review-cycle?

Machine verification is required before the review-cycle to validate inputs and ensure execution accuracy across the provider and model matrix. This governed cycle enforces safety rules, controls lane isolation, and guarantees reliable KPI capture before final review.

Does benchmarking across a provider and model matrix support Brain persistence for artifacts?

Yes, benchmarking across a provider and model matrix supports Brain persistence for artifacts. The workflow captures KPIs and summaries for each lane and the whole suite, persisting these artifacts through the Brain during the governed execution cycle.