mcpField
G
gauge-agents

gauge-agents

github.com/gauge-shTesting & quality018 installsUpdated 1d ago
Repository

About gauge-agents

Gauge Agents is a skill for teams evaluating how coding agents discover and use their products. It uses Gauge CLI sessions to measure agent experience and preferences, inspect evidence, and test documentation or skill changes.

Key features

  1. 01Measure agent product experience
  2. 02Compare agent tool preferences
  3. 03Run controlled CLI sessions
  4. 04Inspect evaluation evidence
  5. 05Test documentation and skill changes

Use cases

  1. 01Assess friction in product usage by coding agents
  2. 02Compare which tools agents choose for tasks
  3. 03Evaluate documentation or skill revisions
  4. 04Set up repository-based PR checks

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/gauge-sh/gauge-skills --skill gauge-agents

After installing, type/gauge-agentsin Claude Code to confirm it loaded.

Bundled filesreferences/github-setup.md

Details

From the project README

Gauge Agents

Install the CLI

If gauge is not installed, install it with Node.js 22.12 or later:

GAUGE_SKIP_SKILL_INSTALL=1 npm install -g @withgauge/cli

The Skill is already installed; skip the CLI’s automatic Skill installation to preserve this selected version and its references.

Load the instructions

Before using Gauge, run:

gauge instructions

Follow the returned instructions for setup and operation. They ship with the CLI and are the source of truth for agent workflows. Use gauge instructions <topic> to reload a specific topic and gauge <command> --help for flags.

Set up repository PR checks

When asked to set up Gauge PR checks, author Git-based eval cases, or connect a repository, read references/github-setup.md. It covers choosing inputs for an npm package, native CLI, docs preview, or Skill, reusing the repository’s CI, authoring advisory Agent Preference cases, selecting relevant file changes, and verifying setup before committing.

What Gauge is

Gauge helps you optimize how coding agents discover and use your product. It runs coding-agent sessions under stable conditions so you can evaluate performance and test improvements.

  • Agent experience: How successfully do agents use your product, where do they struggle, and how can you reduce friction?
  • Agent preference: Which tools do agents select for a given task, and why?

Use the CLI to configure measurements, run sessions, inspect evidence, and test changes to your docs and skills headlessly.

Comments

Markdown supported