Summary

The model receives the prompt (and optional system message). The run uses scorer contains_all with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.

Summarize in exactly 3 short bullet points: API quality improved by contract tests, versioned endpoints, and deprecation notices with migration guides. Breaking changes required architecture review approval.

{
  "expected_contains": [
    "contract tests",
    "versioned",
    "deprecation"
  ]
}