Confidence, recomputed by someone else

Scoring Shipped

Every finding is re-scored 0–100 against an anchored rubric by an actor blind to who found it. Nothing is dropped; the score decides order.

The consultant runs once, over everything, after the pillars finish. It re-scores each finding against a rubric and a false-positive catalog held in team data, writes a verdict and a validated confidence, and marks duplicates. The actor is described under Who does the work → The consultant; the argument for the separation is under Why it works this way.

What the score is used for

  • Ordering the report100every finding, ranked
  • Selecting fixes100--fix recommended reads the disposition
  • Filtering the report0never — see "Why nothing is filtered out"

The empty bar is the design. A score that could remove a finding from the report would be a threshold wearing a rubric.