For Scholars
Full access to the prompts, schemas, and integrity testing framework behind every CFI evaluation. Everything is readable, auditable, and forkable.
Open Resources
System Prompts
7 specialized prompts defining the classifier and six interpretive lenses. Each includes constitutional grounding, scoring rubrics, and output format specifications.
Scoring Schema
JSON Schema defining the output structure for all evaluations, including score cells, confidence levels, relevance ratings, and narrative fields.
Integrity Tests
40 tests verifying scoring symmetry across partisan lines and invariance across different framings of the same action.
Governance Model
The CFI operates under a principle of radical transparency. All system prompts, scoring rubrics, aggregation algorithms, and integrity tests are publicly available for inspection and criticism.
Evaluations are generated by AI using structured prompts. The system is designed to be auditable — any researcher can reproduce evaluations using the published prompts and verify scores against the aggregation pipeline.
Integrity testing includes partisan symmetry pairs (ensuring similar actions by different administrations receive similar scores) and framing invariance tests (ensuring different descriptions of the same action produce consistent results).
Architecture Overview
26
Executive Orders
6
Interpretive Lenses
7
Constitutional Dimensions
1,092
Scoring Cells