SYasJ

LLM Hallucination Detector — AI skill for Claude Code

AI community

Free, open-source LLM hallucination detector.

How to install LLM Hallucination Detector

This entry records only its repository, not the path inside it, so there is no exact command to give. Open SYasJ/llm-hallucination-detector and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What LLM Hallucination Detector does

Free, open-source LLM hallucination detector. See token-level confidence as a heatmap and check every claim against your own sources (RAG). Works with any OpenAI-compatible API, ChatGPT & Claude. Zero dependencies, runs locally.

Alternatives in AI

Karpathy LLM Wiki — Agent Skills-compatible LLM wiki for Claude Code, Cursor, and Codex 618 ★
Browser Search — A skill for AI agents: search the web with SearXNG, browse with Camofox, bypass protections with CloakBrowser 499 ★
Sloptrim — A local detector for AI-writing patterns 196 ★

README

LLM Hallucination Detector

Catch LLM hallucinations before your users do.

Your model says *"Yes, returns are accepted within 90 days"* with **89% confidence**, and your policy says **30**.
LLM Hallucination Detector colours every token by how sure the model was **and** checks every claim against your own sources, so confident mistakes stop slipping through.

[![CI](https://github.com/SYasJ/llm-hallucination-detector/actions/workflows/ci.yml/badge.svg)](https://github.com/SYasJ/llm-hallucination-detector/actions/workflows/ci.yml) [![License: MIT](https://img.shields.io/badge/License-MIT-teal.svg)](LICENSE) ![Python 3.10+](https://img.shields.io/badge/python-3.10%2B-blue) ![Zero dependencies](https://img.shields.io/badge/dependencies-0-brightgreen)

**[▶ Try the live demo](https://syasj.github.io/llm-hallucination-detector/)**: no install, no sign-up, no API key.

[📖 Wiki](https://github.com/SYasJ/llm-hallucination-detector/wiki) · [Quick start](#quick-start) · [Examples](examples/) · [API](#local-http-api) · [Chrome extension](#chrome-extension-chatgpt--claude)

LLM Hallucination Detector: a confident answer about a 90-day return window, highlighted token by token and flagged as contradicting a 30-day policy

Why LLM Hallucination Detector?

Spot the guess. Invented numbers, names, dates and citations usually show up as red, low-probability tokens.
Prove it against your docs. Each claim is marked supported, contradicted or not in source, with the quote that backs it up.
Gate your RAG pipeline. A 0–100 trust score and a JSON export plug straight into evaluation and guardrails.
Works with what you already use. Any OpenAI-compatible API, pasted ChatGPT or Claude answers, and a Chrome extension.
Runs on your machine. Zero dependencies,