# `get_score_history`

Agent-Readiness score over time for a brand.

**Tier:** read · **Rate limit:** ~30/min

## Input

| Param | Type | Required | Notes |
|---|---|---|---|
| `brand` | string | ✅ | Brand name |

## Output

```json
{
  "brand": "Brooklinen",
  "history": [ … ],
  "withheld_runs": [
    { "run_id": "…", "created_at": "…", "blockers": ["…"] }
  ]
}
```

!!!warning `withheld_runs` is not a subset of `history`
Runs whose evidence gates failed are **split out** of `history` into `withheld_runs`.
They are not scores. Charting `history` alone is correct; merging the two arrays and
treating a withheld run as a zero would invent a crash that did not happen.
!!!

## Errors

`NOT_FOUND` when the brand has no score history.

## Reading a trend

Two consecutive points are only comparable when the scoring version and the question set
both match. A movement of 15 points or less is within the
[noise floor](../../audit/reliability.md#the-noise-floor) and should not be reported as
a change.