midnight-fact-check:fact-check-classification

Classify fact-check claims into verification domains and update claims JSON.

37|9|Updated Feb 25, 2026
One-click install
npx skills add https://github.com/devrelaicom/midnight-expert --skill midnight-fact-check-fact-check-classification
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: midnight-fact-check:fact-check-classification
Source: https://github.com/devrelaicom/midnight-expert/tree/main/plugins/midnight-fact-check/skills/fact-check-classification
Command: npx skills add https://github.com/devrelaicom/midnight-expert --skill midnight-fact-check-fact-check-classification

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill removes ambiguity and manual effort from labeling fact-check claims by assigning them to precise verification domains (compact, sdk, zkir, witness), ensuring consistent downstream processing and automated pipeline routing.

Core Features & Use Cases

  • Domain tagging: Identify whether a claim belongs to compact, sdk, zkir, or witness and add the domain to the claim's domains array.
  • Classification merging rules: Merge or append domain tags and update classification.primary_domain and confidence according to explicit overwrite and confidence rules.
  • Cross-domain handling: Mark partial matches with medium confidence and include concise notes explaining cross-domain rationale.
  • Use Case: Run this skill as part of the midnight-fact-check pipeline to prepare claims for later verification agents and reporting.

Quick Start

Classify the claims in the provided claims file for the assigned domain and update each claim's domains and classification fields accordingly.

Frequently Asked Questions about midnight-fact-check:fact-check-classification

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

FAQPage Schema
How do I classify fact-check claims into verification domains?

Fact-check claim classification tags each claim with verification domain labels like compact, sdk, zkir, or witness. The skill reads your claims JSON, appends matched domains to the domains array, and updates classification.primary_domain and confidence according to explicit merging rules.

What verification domains does the Midnight fact-check pipeline support?

The supported verification domains are compact, sdk, zkir, and witness. Claims are tagged with these domain labels to ensure consistent downstream processing and automated agent dispatch during batch operations within the Midnight fact-check pipeline.

How does the skill handle cross-domain claims that partially match multiple domains?

Cross-domain claims are marked with medium confidence and include concise notes explaining the cross-domain rationale. Classification merging rules explicitly define how to append domain tags and update the primary_domain to resolve partial matches.

Can I use this skill to prepare claims JSON for automated agent dispatch?

You can use this skill to prepare claims JSON for automated agent dispatch. It updates each claim's domains and classification fields by writing back to the same file path, enabling individual-domain classification runs for downstream verification agents.

What are the limitations of automated domain tagging for fact-check claims?

A limitation of automated domain tagging is its restriction to the compact, sdk, zkir, and witness domains. Cross-domain partial matches receive medium confidence rather than high confidence, requiring manual review of classification notes for ambiguous claims.