hallucination-detector

Detect fabricated URLs, unsourced statistics, fake citations, and unfounded future claims in agent outputs.

6|Updated Apr 7, 2026
One-click install
npx skills add https://github.com/Kgan01/ghengis-skills --skill hallucination-detector-kgan01
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: hallucination-detector
Source: https://github.com/Kgan01/ghengis-skills/tree/main/plugins/ghengis-skills/skills/hallucination-detector
Command: npx skills add https://github.com/Kgan01/ghengis-skills --skill hallucination-detector-kgan01

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The detector helps identify and flag hallucinations in LLM outputs, catching fabricated URLs, unsourced statistics, fake citations, and unfounded future claims to preserve accuracy.

Core Features & Use Cases

  • Detects and flags fabricated URLs, unsourced statistics, fake citations, and unsubstantiated future claims in agent outputs.
  • Works across tool results and generated text, acting as an on-demand safety layer within reasoning and content pipelines.
  • Useful for QA in research, software development, and knowledge workflows to maintain trust and reduce misinformation.

Quick Start

Invoke the hallucination-detector on the agent output to automatically flag fabricated URLs, unsourced statistics, fake citations, and unfounded future claims before delivery.

Frequently Asked Questions about hallucination-detector

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

FAQPage Schema
How do I detect fabricated URLs and fake citations in LLM outputs?

Detect fabricated URLs and fake citations by applying regex-based detection patterns to agent-generated text, flagging hallucinated content before delivery. This safety layer scans tool results and generated text to preserve accuracy.

What is the best way to flag unsourced statistics in agent responses?

The best way to flag unsourced statistics is to invoke an on-demand hallucination detection layer within reasoning pipelines. It evaluates agent outputs against detection patterns to identify unfounded claims and unsubstantiated data.

Can I use hallucination detection as a safety layer within content pipelines?

Yes, you can use hallucination detection as an on-demand safety layer within content pipelines. It integrates with the reasoning skill-chain scratchpad, reporting structured results under hallucination_detector keys.

How do I check agent outputs for unfounded future claims before delivery?

Check agent outputs for unfounded future claims by running a hallucination detector that applies regex patterns to generated text and tool results. It flags unsubstantiated predictions to maintain trust and reduce misinformation.

Does hallucination detection work on both generated text and tool results?

Yes, hallucination detection works across both agent-generated text and tool results. It functions as a safety layer within reasoning and content pipelines to identify fabricated URLs, fake citations, and unsourced statistics.

When should I not rely on regex-based hallucination detection for factual accuracy?

Regex-based hallucination detection is limited to pattern matching for fabricated URLs, unsourced statistics, and fake citations. It should not be relied upon when deep semantic verification or external fact-checking against live databases is required.