What makes Katalepsis different from a standard AI explanation?

While standard explanations are one-way lectures, Katalepsis is a two-way verification protocol. It asks you questions to ensure you've 'grasped' the logic, rather than just reading a summary.

Can I skip certain parts of the verification?

Absolutely. You can select which categories you want to dive into at the start, and you can exit the protocol or signal that you understand at any time.

What kind of comprehension gaps does it look for?

It specifically looks for gaps in Causality (why it happens), Scope (impact), Sequence (order of operations), and Expectations (behavior vs. assumption).

How do I trigger the Katalepsis protocol?

It activates when you use phrases like 'explain this', 'I don't get it', or 'walk me through what you did' following an AI-generated code change.

Does it track which parts of the code I already understand?

Yes, it categorizes the changes into specific tasks and uses the TaskUpdate tool to mark each section as 'completed' once your understanding is verified.

Katalepsis Comprehension Protocol

Name: Katalepsis Comprehension Protocol
Author: jongwony

byjongwony

•

학습 및 문서화

Verifies user understanding of AI-generated work through structured Socratic dialogue and categorized verification loops.

Katalepsis is an epistemic protocol for Claude Code designed to bridge the gap between AI output and human comprehension. Based on the Stoic concept of "firmly grasping" knowledge, it replaces passive lectures with active verification. When triggered by a user's confusion or request for explanation, the skill categorizes changes into logical domains—such as architecture, logic, or dependencies—and uses the Socratic method to confirm the user truly understands the "how" and "why" of the work. It ensures that the user moves from ungrasped results to verified understanding through structured task tracking and targeted inquiry.

주요 기능

01Socratic Verification: Confirms comprehension through targeted questioning rather than passive explanation.

02Change Categorization: Automatically groups AI work into themes like Architecture, Bug Fixes, and Logic for structured learning.

03Task-Based Progress: Integrates with internal task tools to track the completion of learning objectives.

04Comprehension Gap Detection: Identifies specific misunderstandings in causality, scope, sequence, or expectations.

05Multi-Level Intensity: Adjusts the depth of verification based on the complexity of the changes and user familiarity.

0643 GitHub stars

사용 사례

01Confirming a firm grasp of security-critical logic or breaking changes before finalizing a pull request.

02Onboarding developers to unfamiliar sections of a codebase where AI has implemented new functionality.

03Reviewing complex architectural refactors to ensure the underlying design patterns are fully understood.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jongwony/epistemic-protocols katalepsis

For use in Claude.ai and ChatGPT

Download Skill