Verify factual claims in documents using multi-LLM triage and source-grounded verification.
npx skills add https://github.com/machug/fact-checker --skill fact-check-machug
Please help me install this Agent Skill. Skill: fact-check Source: https://github.com/machug/fact-checker/tree/main/skills/fact-check Command: npx skills add https://github.com/machug/fact-checker --skill fact-check-machug
Systematically verify factual claims in documents using a hybrid approach: multi-LLM triage and source-grounded verification.
Provide a document and run the full pipeline to extract claims, triage with multiple models, verify against authoritative sources, and generate a report.
High-intent search queries and answers about installing and using this skill.
You can verify factual claims in technical documentation automatically by using a multi-LLM triage pipeline that extracts statements and cross-checks them against authoritative sources. This systematically flags uncertainties in pricing, dates, and compliance materials.
Multi-LLM triage for document fact-checking is a process where extracted claims are sent to multiple language models for cross-checking and uncertainty flagging. It identifies potentially inaccurate information before deep verification against external sources.
Yes, you need the litellm library to verify document facts with multiple models, as it manages the model calls required for the triage phase. Optional external providers and an MCP-backed source registry are also used for authoritative verifications.
To check pricing and licensing accuracy in Azure documentation before publishing, run a full fact-checking pipeline that extracts claims, triages them with multiple models, and verifies them against credible external sources to generate a structured accuracy report.
The best way to source-ground verification for software documentation claims is to query MCP servers and credible external providers during the verification phase. This ensures claims about capabilities and dates are checked against authoritative registries rather than relying solely on model inference.
A limitation of using multi-LLM triage for compliance and governance fact-checking is its reliance on the availability of MCP-backed source registries and optional external providers. Verification accuracy depends on the coverage and credibility of these external authoritative sources.
Miosa-osaExtract atomic claims with provenance and confidence from source materials.
ValorInvestigatorExtract claims, gather evidence, and document corrections for journalism fact-checking.
thisrohanguptaAutomate Harness Connector creation and testing via Harness MCP v2.
framerslabVerify AI-generated responses against source documents using NLI entailment and LLM assessment.
