What files does this skill search?

The skill searches through your local data directories, including meeting transcripts, markdown notes, and project documentation to find textual matches for claims.

How does Grounded Query prevent AI hallucinations?

It forces the AI to extract its own assertions and then search through your local data files for specific evidence using a multi-step verification workflow before finalizing the output.

Does this skill work with source code or just documentation?

While it can search any text file, it is optimized for verifying assertions about decisions, status, and plans typically found in transcripts and project documents.

What happens if no evidence is found for a claim?

The skill will mark the status as 'NOT FOUND' in the evidence table and append a note warning that confidence is lower for those specific unverified items.

Grounded Query Fact-Checker

Name: Grounded Query Fact-Checker
Author: my-entourage

bymy-entourage

•

Productivity & Workflow

Verifies factual claims against source documents to ensure accuracy and prevent AI hallucinations when querying project data.

About

Grounded Query is a specialized skill designed to improve the reliability of AI-generated responses by cross-referencing assertions with local data files. It automatically identifies factual claims—such as project statuses, meeting decisions, or specific timelines—and systematically searches transcripts and documentation to find supporting evidence. By generating a structured evidence table that classifies findings as Supported, Partial, or Not Found, it provides a transparent audit trail that ensures your AI agent remains grounded in reality rather than making assumptions.

Key Features

2 GitHub stars
Contextual evidence searching using Grep and Read commands
Hallucination prevention for meeting transcripts and project logs
Evidence classification (SUPPORTED, PARTIAL, NOT FOUND)
Automated factual claim extraction from AI responses
Standardized evidence tables with source file pathing

Use Cases

Cross-referencing project status claims against development documentation
Summarizing complex project history with verifiable citations
Verifying project decisions and action items from meeting transcripts

About

Key Features

2 GitHub stars
Contextual evidence searching using Grep and Read commands
Hallucination prevention for meeting transcripts and project logs
Evidence classification (SUPPORTED, PARTIAL, NOT FOUND)
Automated factual claim extraction from AI responses
Standardized evidence tables with source file pathing

Use Cases

Cross-referencing project status claims against development documentation
Summarizing complex project history with verifiable citations
Verifying project decisions and action items from meeting transcripts