The spec

The deliverable

The rubric — every criterion is checked twice Edit lines freely; one criterion per line.

Preset

Saved rubrics

Context notes (optional)

The verdict

Since the previous run of this spec + rubric

Criterion	Previous	Now	Change

Every criterion, both reviewers Rows where B and C disagree are badged — a disagreement means one reviewer caught something the other missed.

Criterion	Reviewer B	Reviewer C

Reviewer B

Read the spec first, then hunted the deliverable for each requirement.

Reviewer C

Read the deliverable first, then checked every claim back against the spec.

Fix plan — merged and deduplicated

Verdict Gate is an automated review, not a substitute for the checks your domain legally requires. Independence reduces shared blind spots; it does not eliminate them — for critical output, add a human spot-check.

Result

How it works

Nothing to hand? — a launch email that quietly breaks its own spec in five places a single self-review tends to miss.

Paste spec and deliverable

Pick a rubric preset or write your own criteria — each one gets an objective pass/fail from both reviewers.

Two independent reviews

Reviewer B walks the spec requirement by requirement; Reviewer C walks the deliverable claim by claim. Neither sees the other’s assessment — a shared blind spot is exactly what this breaks.

Both must pass

NICE ships. NAUGHTY gets a deduplicated fix plan — blockers first, tagged by who caught what — plus an agreement metric that tells you when the rubric is too loose. Copy it as a tracker checklist, copy it as a ready-made fix prompt with your text attached, or export the whole thing as a dated Markdown record. Then re-check the revision and see, criterion by criterion, what you actually fixed.

Derived from the @affaan-m/santa-method skill on SkillSafe.