Value a pool of candidate assets

Measure which documents, examples, and policies improve your agent , with fresh controls, confidence intervals, and per-dollar ranking.

kno value measures the marginal effect of every candidate asset you are considering adding.

Prerequisites

A completed baseline and a pool file. Pools are JSONL, CSV, or Markdown:

{
  "id": "refund-policy-v3",
  "content": "Refunds are processed within 5 business days.",
  "kind": "knowledge"
}

kno value --evals cases.jsonl --pool pool.jsonl --baseline-run-id <run id>

Kno plans the measurements, injects each asset into the slices it could affect, and re-measures against fresh controls.

ASSET              DELTA (95% CI, positive = goal dir)  CONTROL     NOTE
brand-guide        +0.0000 [-0.1260, +0.1260]  low -0.1938 (underpowered)

DELTA, paired per-case difference with a 95% confidence interval. A delta without its interval is not reported at all.
CONTROL, the same cases measured without the asset. Regression is measured separately from improvement.
underpowered, too few cases for a meaningful interval; Kno reports a blank rather than a bad number.
delta-per-cost, ranking is per dollar, not per point. A tiny asset that improves 5% can beat a giant document that improves 6%.

Nothing in this stage reads the holdout sealed at baseline. That separation is the reason the numbers mean anything later.