Which reasoning methods does this framework support?

The framework supports Chain-of-Thought (CoT), Tree-of-Thoughts (ToT), Graph-of-Thoughts (GoT), and Self-Consistency, routing to the optimal method based on task complexity.

What is the Producer-Critic Orchestrator?

It is a design pattern provided by the framework that separates generation (Producer) from evaluation (Critic) to create iterative refinement loops in agent workflows.

How does it help reduce AI hallucinations?

It implements a Self-Verification layer that uses Chain-of-Verification (CoVe) and backward verification to validate factual claims and logical steps before output delivery.

Can this skill detect biases in AI output?

Yes, it includes a dedicated Bias Detector that identifies common cognitive biases like confirmation bias, anchoring, and framing effects within the reasoning process.

Is this skill suitable for production-grade agents?

Absolutely. It is designed to provide the architectural components necessary for building reliable, transparent, and self-aware agents for complex production environments.

AI Critical Thinking Framework

Name: AI Critical Thinking Framework
Author: rfxlamia

byrfxlamia

•

データサイエンスとML

Architects sophisticated AI agents with advanced reasoning capabilities, metacognitive monitoring, and automated self-correction cycles.

The AI Critical Thinking Framework (FCT) is a specialized skill for Claude Code designed to help developers build high-reliability AI agents capable of deep, transparent reasoning. It provides a structured architectural blueprint for implementing state-of-the-art cognitive patterns such as Tree-of-Thoughts (ToT), Graph-of-Thoughts (GoT), and Chain-of-Verification (CoVe). By integrating metacognitive monitoring and bias detection, this framework allows agents to self-assess their confidence, detect logical fallacies, and choose the most efficient reasoning path for complex problem-solving. It is an essential tool for any project requiring autonomous agents that must minimize hallucinations and maintain rigorous logical consistency.

主な機能

01Metacognitive monitoring with real-time confidence scoring

02Self-verification loops using Chain-of-Verification (CoVe) techniques

03Producer-Critic orchestration for multi-step agent refinement

0442 GitHub stars

05Intelligent Reasoning Router for CoT, ToT, and GoT selection

06Cognitive bias detection for mitigation of logical errors

ユースケース

01Implementing complex decision-making workflows with transparent reasoning chains

02Reducing hallucinations in RAG systems through structured output verification

03Building autonomous agents that require multi-step planning and self-correction

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rfxlamia/claude-skillkit framework-critical-thinking

For use in Claude.ai and ChatGPT

Download Skill