What is 'Corollary Discharge' in this context?

Corollary Discharge is the control mechanism that suppresses expected, self-generated signals so that only unexpected anomalies or external threats reach the user's attention.

What is the main purpose of the Reafference & Corollary Discharge skill?

It uses neuroscience principles to help Claude distinguish its own generated actions from external changes, ensuring the system remains secure and behaviorally consistent.

How does the skill identify a security threat?

It compares a predicted 'Efference Copy' of an action with the actual observed history in ~/.claude/history.jsonl; any significant mismatch between the two triggers a critical threat alert.

Does this skill require any external databases?

Yes, it integrates with DuckDB to store and manage 1,260+ records across multiple tables for signal analysis, error tracking, and historical verification.

Reafference Behavioral Verification

Name: Reafference Behavioral Verification
Author: plurigrid

byplurigrid

•

セキュリティとテスト

Implements von Holst’s reafference theory to distinguish self-generated AI interactions from external anomalies and security threats.

This skill brings advanced neuroscience principles to the Claude environment by implementing a four-layer signal processing system based on von Holst’s reafference theory. By generating a deterministic 'Efference Copy' of its own actions and comparing them against observed interaction history, the skill can automatically suppress expected feedback and highlight genuine deviations or external interference. This is an essential tool for developers and researchers building high-integrity autonomous systems who need to verify that an AI's state changes are self-originated rather than caused by unauthorized external actors or system corruption.

主な機能

01Real-time sensory reafference matching against Claude history logs

02Intelligent signal suppression to eliminate self-generated noise

037 GitHub stars

04Deterministic Efference Copy generation using SHA-256 hashing

05Automated error signal computation and threat level classification

06Comprehensive DuckDB integration for persistent signal and threat logging

ユースケース

01Verifying the behavioral integrity of long-running autonomous coding sessions

02Detecting unauthorized external modifications to an AI's interaction history

03Filtering out predictable system feedback to focus on unexpected anomalies

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add plurigrid/asi reafference-corollary-discharge

For use in Claude.ai and ChatGPT

Download Skill