Agent Evaluation

Über diesen Skill

Build evaluation datasets, run and grade agent traces, and investigate failures.

Warum wir es empfehlen

Covers dataset schema, metrics, judge configuration and an iterative eval-and-fix workflow.

Wobei es hilft

• Covers dataset schema, metrics, judge configuration and an iterative eval-and-fix workflow.

Scope and use

Covers dataset schema, metrics, judge configuration and an iterative eval-and-fix workflow.

Install the suite together: this task refers to the shared workflow and other Skills. Actual cloud operations depend on your project configuration and require the appropriate authorization.

Vor der Nutzung

• The Skills work with a coding agent and the bundled references. Task execution uses agents-cli; upstream lists Python 3.11+, uv and Node.js. Local model calls require configured provider access; cloud deployment needs a Google Cloud project and may incur charges.