Test fixture
Code transformation while preserving behavior and intent.
The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.
Return JSON only with keys plan, refactor, tests. UI code directly imports database queries and normalizes rows inline. Refactor a data access boundary that separates fetching, mapping, and rendering.
{
"metrics": {
"intent": {
"checks": [
{
"contains": [
"data access boundary"
]
},
{
"contains": [
"fetching"
]
},
{
"contains": [
"mapping"
]
}
]
},
"visible": {
"checks": [
{
"contains": [
"UI code"
]
},
{
"contains": [
"database queries"
]
},
{
"contains": [
"rendering"
]
}
]
},
"hidden": {
"checks": [
{
"contains": [
"same props"
]
},
{
"contains": [
"mocked tests"
]
},
{
"contains": [
"no direct imports"
]
}
]
}
}
}temperature
0
max_tokens
560
timeout (s)
120
type
scored
file
refactor-data-access-boundary.json