How does the threat detection mechanism work?

The skill uses a Comparator to calculate the distance between an 'Efference Copy' (expected result) and 'Sensory Reafference' (actual result). If the error magnitude exceeds a specific threshold, it escalates the signal as a WARNING or CRITICAL threat.

Can this skill be used for security auditing?

Yes, it is specifically designed for behavioral verification and can identify when an agent's history has been tampered with or when an external force has influenced the model's output patterns.

What is reafference theory in the context of AI?

Reafference theory is a neuroscience principle where the system predicts its own sensory feedback. In this skill, it allows Claude to distinguish its own generated actions from external changes by comparing predicted hashes against actual results.

What is an Efference Copy in this skill?

An Efference Copy is a deterministic prediction of an interaction's outcome, generated by hashing interaction content into a specific color space, ensuring identical inputs always produce the same prediction for verification.

Reafference & Corollary Discharge

Name: Reafference & Corollary Discharge
Author: plurigrid

byplurigrid

•

安全与测试

Distinguishes self-generated AI interactions from external anomalies using von Holst's neuroscience principles and predictive signal processing.

This skill implements a sophisticated behavioral verification system based on von Holst’s reafference theory to enhance AI reliability and security. By generating a deterministic 'efference copy' (prediction) for every interaction and comparing it against actual sensory data (history logs), the system can mathematically distinguish between intended actions and external interference. It provides a robust framework for threat detection, signal suppression, and behavioral consistency analysis, making it ideal for developers building autonomous agents that require high-integrity execution environments.

主要功能

01Corollary Discharge mechanism for automated signal suppression and anomaly escalation

02Automated Comparator logic for calculating signal error and threat levels (SAFE/WARNING/CRITICAL)

038 GitHub stars

04Real-time Sensory Reafference matching against Claude interaction history

05Persistent signal analysis and seed recovery integrated with DuckDB

06Deterministic Efference Copy generation using SHA-256 to color-index mapping

使用场景

01Detecting external prompt injections or unauthorized command modifications in real-time

02Analyzing system drift and interaction patterns for forensic debugging and auditing

03Verifying the behavioral integrity of long-running autonomous agent sessions

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