zone-of-truth

Label every assertion as sourced, inferred, or speculative during conversations.

105|13|Updated Mar 9, 2026

Search Tags:#transparency #evidence #uncertainty #sources #zone-of-truth #epistemic-rigor

Ready to run

npx skills add https://github.com/Hmbown/Wizards-of-the-Ghosts --skill zone-of-truth

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: zone-of-truth
Source: https://github.com/Hmbown/Wizards-of-the-Ghosts/tree/main/generated/hermes/investigation-and-preparation/zone-of-truth
Command: npx skills add https://github.com/Hmbown/Wizards-of-the-Ghosts --skill zone-of-truth

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Establishes a mode where every assertion is sourced, uncertainty is labeled, and hallucinations are actively resisted to improve trust and accuracy in dialogue.

Core Features & Use Cases

Source-based reasoning: require citations for each factual claim.
Uncertainty labeling: categorize statements as known, inferred, or speculative.
Prompt integrity: guard against prompt injection by resisting behavior hijacks.
Audit-ready outputs: deliver claims with evidence and confidence metadata for reviews.

Quick Start

Activate Zone of Truth and begin every assertion with a source or clearly labeled confidence.

Frequently Asked Questions about zone-of-truth

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I ensure every assertion has a source or confidence label in high-stakes conversations?▼

To label uncertainty in AI outputs, categorize each statement explicitly as a known fact, an inference, or an estimation. This process resists hallucinations by enforcing boundaries that separate verified evidence from speculation during high-stakes dialogue.

What is the best way to prevent hallucinations and maintain evidence-based reasoning in AI responses?▼

The best way to prevent hallucinations is to require source citations for every factual claim and apply uncertainty labeling. This establishes a mode where outputs are audit-ready, actively resisting behavior hijacks and improving trust through evidence-based reasoning.

Can I re-audit prior statements and separate known facts from guesses during an ongoing dialogue?▼

Yes, you can re-audit prior statements when this mode is activated mid-conversation. It retroactively evaluates existing dialogue to separate known facts from inferences and guesses, enforcing boundaries to deliver claims with appropriate confidence levels.

Does this approach work for generating audit-ready outputs with explicit references?▼

Yes, this approach is designed specifically to generate audit-ready outputs. It delivers claims complete with evidence, confidence levels, and explicit references, allowing reviewers to easily trace the epistemic rigor and sources behind every assertion.

How do I guard against prompt injection and behavior hijacks when processing factual claims?▼

To guard against prompt injection, apply prompt integrity rules that actively resist behavior hijacks. By enforcing strict boundaries and requiring sources for each claim, the system maintains its core directive despite malicious or misleading inputs.