Can I run verification without a full context load?

Yes, the initial static checks run via zero-context scripts to save on token usage while still catching structural and syntax errors.

How are findings categorized?

Issues are labeled as Critical (potential failures), Concern (structural risks), or Note (stylistic observations) to help you prioritize improvements.

Does this skill automatically block my commits?

No. Following the philosophy of user agency, it identifies issues and lets you decide whether to fix them, review them, or proceed with the commit anyway.

What kind of expert reviews does it perform?

It spawns parallel subagents to analyze Type Design (soundness), Specification Clarity (interpretability), and Claude Code Ecosystem Patterns (best practices).

What is the primary goal of the /verify command?

It surfaces potential issues in protocol definitions—such as schema errors, type inconsistencies, or instruction ambiguity—before you commit changes to ensure high-quality AI behavior.

Protocol Verification

Name: Protocol Verification
Author: jongwony

byjongwony

•

Security & Testing

Validates the consistency and quality of epistemic protocols through automated static analysis and multi-perspective expert reviews before code commits.

The Protocol Verification skill provides a robust validation layer for Claude Code plugins, focusing on the integrity of epistemic protocols. It executes automated static checks for JSON schema and notation consistency while simultaneously deploying parallel LLM subagents to audit type design, specification clarity, and ecosystem alignment. Designed around the principle of 'surface, don't enforce,' it empowers users to identify critical flaws, logical inconsistencies, and UX friction points in complex AI instructions without blocking development workflows or sacrificing user agency.

Key Features

01Automated static JSON schema and Unicode notation validation

02Real-time synthesis of findings categorized by Critical, Concern, and Note severity

03Specialized Claude Code Ecosystem review to filter context-specific false positives

04Interactive decision-making via AskUserQuestion for commit gating

05Parallel multi-agent expert reviews for type design and specification clarity

0643 GitHub stars

Use Cases

01Validating complex state machines and type signatures in protocol definitions before a git commit

02Auditing cross-reference integrity between project documentation and source files

03Ensuring LLM instruction clarity and tool-binding accuracy in new Claude Code skills

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jongwony/epistemic-protocols verify

For use in Claude.ai and ChatGPT

Download Skill