Skip to main content
Open Skills
agent-platform-eval-flywheelby GoogleOfficialDevelopmentGitHub stars: 20.6k

Agent Platform Eval Flywheel skill: what it does and how to install it

Evaluates and improves AI models and agents on Google Cloud with synthetic scenarios, rubric metrics, failure analysis and result comparison.

Summary generated from the skill's documentation.

Install

$ npx skills add google/skills --skill agent-platform-eval-flywheel

Run it in a terminal. If your agent is already running, start a new session so it picks the skill up.

About this skill

What it does. Guides the Agent Platform GenAI Evaluation SDK through dataset preparation, inference, grading, failure analysis and iterative optimisation. It supports session traces, DataFrames, synthetic scenarios, built-in and custom metrics, persisted JSON and HTML reports, and before-and-after comparisons.

When to use it. For evaluating Google Cloud models or agents, selecting metrics, analysing failures or checking whether fixes improve results. It does not cover fine-tuning or general production deployment.

History

Repo stars

20.6kAbout +2.3k since 9 Jul 2026

Before 1 Oct 2026 the curve is estimated from public event data.

Stars are counted for the whole repository, which holds 36 skills.

Show as a table
Repo stars by day
DateRepo stars
1 Oct 202620,557
24 Sept 2026 (estimated)20,498
17 Sept 2026 (estimated)20,452
10 Sept 2026 (estimated)19,878
3 Sept 2026 (estimated)18,833
27 Aug 2026 (estimated)18,776
20 Aug 2026 (estimated)18,765
13 Aug 2026 (estimated)18,765
6 Aug 2026 (estimated)18,294
30 Jul 2026 (estimated)18,271
23 Jul 2026 (estimated)18,271
16 Jul 2026 (estimated)18,260
9 Jul 2026 (estimated)18,248

Installs

7.1k

Tracking since . A chart appears once there are 7 days of data.

Installs via skills.sh

Similar skills