How does this skill prevent AI hallucinations?

It enforces a rule where every statement must be tagged as [Confirmed], [Inferred], or [Assumption] based on explicit code evidence and file references, preventing the AI from stating guesses as facts.

What is the difference between Certainty Tags and Derivation Tags?

Certainty Tags are used for analyzing existing code (e.g., [Confirmed]), while Derivation Tags are used when generating new content to track if a requirement is direct or AI-derived.

How do I set up the tag language preference?

The skill checks for a 'Certainty Tag Language' section in your CONTRIBUTING.md file. If not found, it will default to English or ask for your preference during the first run.

Can I use this for non-English projects?

Yes, the skill supports configuration via the project's CONTRIBUTING.md file to use either English or Chinese (繁體中文) certainty tags.

AI Collaboration Standards

Name: AI Collaboration Standards
Author: ValorVie

byValorVie

•

コラボレーションツール

Standardizes AI communication using certainty tags and source citations to eliminate hallucinations and improve decision-making accuracy.

This skill implements a rigorous framework for AI-human collaboration by enforcing evidence-based reasoning and transparent certainty levels. It introduces a unified tagging system—distinguishing between confirmed facts, logical inferences, and assumptions—while requiring file-level source citations for every claim. By mandating that the AI provides clear recommendations and identifies information gaps, it transforms Claude into a more reliable and accountable pair programmer, making it ideal for code reviews, architectural analysis, and complex refactoring tasks where accuracy is paramount.

主な機能

01Strict Source Citation (File paths and line numbers required)

02Mandatory Recommendation Logic for presented options

03Multi-language Configuration Support (English/Chinese)

043 GitHub stars

05Derivation Tracking (Source vs. AI-Generated content)

06Unified Certainty Tagging (Confirmed, Inferred, Assumption, Unknown)

ユースケース

01Generating technical specifications from requirements while tracking derivations

02Analyzing complex bugs where distinguishing facts from guesses is critical

03Performing objective code reviews and reverse engineering legacy systems

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add valorvie/custom-skills ai-collaboration-standards

For use in Claude.ai and ChatGPT

Download Skill