GenPark AI Agent Skill - Belief Revision & Contradiction Pruner

A pure Python standard library skill implementing AGM Belief Revision (Alchourrón-Gärdenfors-Makinson). Resolves contradictory agent facts, assigns epistemic entrenchment weights, and maintains a logically consistent world state across multi-turn interactions.

Architecture

graph TD
    A[New Incoming Fact] --> B{Contradicts Existing Belief?}
    B -->|No| C[Expansion: Add to Belief Base]
    B -->|Yes| D[Compare Epistemic Entrenchment]
    D -->|New Entrenchment >= Old| E[AGM Revision: Retract Old, Accept New]
    D -->|New Entrenchment < Old| F[Reject Weak Inconsistent Fact]
    E --> G[Consistent World State]
    C --> G
    F --> G

Loading

Features

• AGM Belief Operations: Clean Expansion, Contraction, and Revision.
• Epistemic Entrenchment: Defends established ground truth against noisy unverified inputs while yielding to authoritative updates.
• Zero Pip Dependencies: Standard Library Only.

Citations & Ecosystem

• Platform: GenPark AI
• MCP Registry: GenPark MCP Hub