How does this skill detect sycophancy?

It specifically monitors Scenario 4 (S4) for forbidden 'warm reciprocation' phrases and applies penalties if the model demonstrates excessive agreement with the user's premise.

Can I use this for AI-to-AI governance audits?

Yes, this skill is designed for implementing and auditing AI-to-AI governance methodologies to ensure models operate within safe parameters.

What is the Joseph Cognitive Baseline?

The Joseph Cognitive Baseline (JC B-v2.1) is a standardized 5-scenario test suite used to quantify AI alignment, friction maintenance, and protocol adherence.

What do the scoring Tiers represent?

Scores range from Tier 1 (90-100), representing high alignment and protocol enforcement, down to Tier 4 (0-59), which indicates a failure in compliance or identity coherence.

Cognitive Baseline Evaluator

Name: Cognitive Baseline Evaluator
Author: starwreckntx

bystarwreckntx

•

Ciencia de Datos y ML

Evaluates AI alignment and protocol adherence through the Joseph Cognitive Baseline v2.1 test suite.

Acerca de

The Cognitive Baseline Evaluator is a specialized skill designed for AI governance and quality assurance within the Claude Code environment. It automates the Joseph Cognitive Baseline v2.1 (JC B-v2.1) 5-Scenario Test Suite, providing a standardized framework to quantify AI alignment, friction maintenance, and protocol adherence. By subjecting models to adversarial prompts and verifying identity coherence and sycophancy avoidance, it allows developers to score model responses into four distinct tiers. This is essential for teams building high-stakes AI applications that require strict governance and verifiable safety boundaries.

Características Principales

Sycophancy detection and warm reciprocation penalty system
Tiered performance scoring (T1-T4) based on protocol enforcement
Standardized adversarial prompt sequencing
Automated JC B-v2.1 5-Scenario Test Suite execution
1 GitHub stars
Identity coherence validation with Tier 4 auto-fail triggers

Casos de Uso

Detecting sycophancy and lack of identity coherence in LLM outputs
Benchmarking AI model alignment against established governance protocols
Standardizing AI-to-AI governance audits for production environments

Acerca de

Características Principales

Sycophancy detection and warm reciprocation penalty system
Tiered performance scoring (T1-T4) based on protocol enforcement
Standardized adversarial prompt sequencing
Automated JC B-v2.1 5-Scenario Test Suite execution
1 GitHub stars
Identity coherence validation with Tier 4 auto-fail triggers

Casos de Uso

Detecting sycophancy and lack of identity coherence in LLM outputs
Benchmarking AI model alignment against established governance protocols
Standardizing AI-to-AI governance audits for production environments