---
name: coherence-scale-author
description: Create or revise a valid Coherence Dimension and numeric/ordinal Scale. Use when a Realm or entity needs a measurable construct, bipolar poles, landmarks, evidence rules, confidence gates, valid uses, and calibration cases.
---

# Coherence Scale Author

Create an observable, falsifiable construct. Do not disguise a value, diagnosis, identity, or reputation judgment as measurement.

## Required inputs

Require purpose/decision supported, namespace and publishing Authority, entity/scope applicability, behavioral construct, contexts, candidate poles, available evidence, consequence, cultural/accessibility limits, Consent/privacy, and boundary relationship. Load dimension/scale schemas and `../../docs/DIMENSIONS-AND-SCALES.md`.

## Procedure

1. State one construct in behavioral terms. Split compound concepts.
2. Write negative and positive poles so neither is inherently moralized. Define optional midpoint.
3. Choose a numeric or ordinal range. Prefer `-100…+100` when direction/distance help; use another range when justified.
4. Add ordered landmarks with observable anchors, not personality labels.
5. Define indicators, counter-indicators, alternative causes/confounds, falsifiers, applicable contexts/entities, and eligible observers.
6. Set minimum evidence count, context count, half-life, consequence-specific confidence gate, and method confidence ceiling.
7. State permitted and prohibited uses, cultural/context limits, privacy, and social-scoring prohibition.
8. Keep local target/tolerance/zones out of the portable dimension; profiles provide them. Reference absolute boundaries only through governed policies.
9. Create rehearsal cases: both poles, midpoint, non-zero target, contradictory evidence, low confidence, cultural/context difference, and boundary distinction.
10. Validate both JSON objects before returning and mark status `draft` or `rehearsal` until calibration.

## Output

Return one schema-valid `CoherenceScale`, one schema-valid `CoherenceDimension`, a validity rationale, a calibration/evaluation set, unresolved risks, and required publication Authority.

## Constraints and escalation

- Never publish a clinical, diagnostic, moral-worth, intelligence, trustworthiness, or global human score without a separately legitimate validated system; default action is to redesign or reject.
- “Trust” must be decomposed into relationship-specific observable behavior such as disclosure, verification, or reliability.
- Ask when poles are ambiguous or the same behavior has opposite plausible causes.
- Escalate system namespace publication to Kiduna governance and Realm namespace publication to Realm governance.

## Example

Replace “Empowerment” with separate Initiative and Agency-support dimensions. Define observable behavior and prohibited human-ranking use rather than scoring whether a person is empowered.