Waza Runner

Skill Eval Runner

Evaluate Agent Skills like you evaluate AI Agents

This skill runs evaluations on other skills to measure their effectiveness using the same patterns that power AI agent evaluations.

When to Use

• Running quality evaluations on a skill
• Testing if a skill triggers on correct prompts
• Measuring skill behavior quality
• Generating eval reports for CI/CD

Commands

Run Evals

Run evals on <skill-name>

Initialize Eval Suite

Create evals for <skill-name>

Generate Report

Generate eval report for <skill-name>

Workflow

1. Check for Eval Suite: Look for eval.yaml in the skill directory
2. Load Tasks: Parse task definitions from tasks/*.yaml
3. Execute: Run each task through the configured graders
4. Report: Output results in JSON or Markdown format

Metrics Measured

Grader Types

• Code Graders: Deterministic assertions, regex matching
• LLM Graders: Model-as-judge with configurable rubrics
• Human Graders: Manual review workflow

Example Usage

Running Evals

# From CLI
waza run ./my-skill/eval.yaml

# Output to file
waza run ./my-skill/eval.yaml -o results.json

Interpreting Results

{
  "summary": {
    "pass_rate": 0.85,
    "composite_score": 0.82
  },
  "metrics": {
    "task_completion": { "score": 0.9, "passed": true },
    "trigger_accuracy": { "score": 0.95, "passed": true }
  }

References

• Eval Specification (opens in a new tab) - Full eval.yaml schema
• Writing Tasks (opens in a new tab) - Task definition guide
• Grader Reference (opens in a new tab) - Available graders