evals-creator

Generate step and scenario eval definitions from intent.yaml and skill contracts.

3|Updated Jan 25, 2026
One-click install
npx skills add https://github.com/kapilvirenahuja/garura --skill evals-creator
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: evals-creator
Source: https://github.com/kapilvirenahuja/garura/tree/main/core/components/skills/evals-creator
Command: npx skills add https://github.com/kapilvirenahuja/garura --skill evals-creator

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill eliminates the manual, error-prone effort of writing and keeping eval definitions in sync with evolving intents and skill contracts, by generating structured step and scenario evals from the source documents.

Core Features & Use Cases

  • Step eval generation: Creates step evals mapped to intent failure conditions and constraint-driven artifact-verifiable rules, each with concrete, observable pass/fail criteria.
  • Scenario eval generation: Builds E2E scenario evals from expectation success scenarios (or legacy intent scenarios) so play-level outcomes remain consistent and traceable.
  • Traceable, binary checks: Ensures every eval is traceable to its source (failure condition, constraint, or scenario) and grounded in observable artifact state.
  • Coverage reporting: Produces a coverage summary indicating which failure conditions, constraints, and scenarios are covered or uncovered.

Quick Start

Use evals-creator to generate updated step and scenario eval YAML by providing the intent_path, skill_contracts, output_path, and (optionally) expectation_path.

Frequently Asked Questions about evals-creator

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

FAQPage Schema
How do I generate scenario evals from intent YAML files?

To generate scenario evals from intent YAML, provide the intent_path, skill_contracts, output_path, and optionally expectation_path to produce structured, traceable step and scenario eval definitions mapped to failure conditions.

What is artifact verification in agentic workflow evals?

Artifact verification in agentic workflow evals ensures every generated check is grounded in observable artifact state, enforcing binary, traceable pass/fail criteria mapped directly to specific constraints and failure conditions.

How do I keep eval definitions in sync with evolving skill contracts?

Keep eval definitions in sync by generating structured step and scenario eval YAML directly from source intent and skill contracts, eliminating manual effort and ensuring checks remain traceable to evolving constraints.

Can I generate coverage reports for uncovered failure conditions?

Yes, you can generate coverage reports. The process produces a coverage summary indicating which failure conditions, constraints, and scenarios are currently covered or uncovered by the generated eval definitions.

Do I need an expectation.yaml file to generate step evals?

You do not need an expectation.yaml file to generate step evals. The process generates step evals from intent.yaml and skill contracts, using expectation.yaml optionally to build E2E scenario evals.

How do I map failure conditions to observable binary checks?

Map failure conditions to observable binary checks by parsing constraints and failure_conditions from source documents, generating step evals that enforce artifact-verifiable constraint handling with concrete pass/fail criteria.