aget-check-facts

Verify factual claims through a three-pass provenance and logic protocol.

1|Updated Oct 20, 2025
One-click install
npx skills add https://github.com/aget-framework/template-spec-engineer-aget --skill aget-check-facts
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: aget-check-facts
Source: https://github.com/aget-framework/template-spec-engineer-aget/tree/main/.claude/skills/aget-check-facts
Command: npx skills add https://github.com/aget-framework/template-spec-engineer-aget --skill aget-check-facts

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prevents hallucinated or weakly sourced statements from spreading by verifying factual claims with traceable evidence and uncertainty-aware reporting.

Core Features & Use Cases

  • Stratified 3-pass verification: data verification, logic verification, then propagation assessment to decide how much confidence to apply.
  • Read-only factual integrity: performs verification without modifying KBs, session files, or L-doc artifacts.
  • Multi-scope operation: supports self-checking recent conversation assertions, targeted claim checking, and artifact/file-based verification.
  • Uncertainty downgrades: marks any unverified outcomes using explicit uncertainty language to reduce false confidence.
  • Anti-confabulation micro-gate: applies provenance, authority, and recency checks to future assertions related to verified topics.

Quick Start

Run /aget-check-facts with a specific claim to verify it using source traceability, then receive a PASS/WARN/CRITICAL style report with downgraded uncertainty where needed.

Frequently Asked Questions about aget-check-facts

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

FAQPage Schema
How do I verify factual claims and trace entities to authoritative sources?

Fact verification traces each entity, date, count, and identifier to authoritative sources, validating inferences and downstream propagation risk. A deterministic three-pass protocol checks data, logic, and propagation to return PASS, WARN, or CRITICAL reports with uncertainty downgrades.

How does uncertainty management work when provenance cannot be fully validated?

Uncertainty management applies explicit uncertainty downgrades when provenance or logic cannot be fully validated. This marks unverified outcomes using uncertainty language to reduce false confidence and prevent weakly sourced statements from propagating.

Can I validate claims contained in session artifacts or L-docs without modifying them?

Read-only audits validate claims in session artifacts or L-docs without modifying knowledge bases, session files, or artifacts. The verification operates across multiple scopes including self-checking recent conversation assertions and targeted claim checking.

What is the best way to prevent hallucinated statements from spreading in conversations?

Anti-confabulation micro-gates prevent hallucinated statements from spreading by applying provenance, authority, and recency checks to assertions. Verified topics receive confidence levels based on source traceability and logic validation outcomes.

What are the limitations of read-only fact verification for logic validation?

Read-only fact verification cannot modify knowledge bases or session files, and applies uncertainty downgrades when provenance or logic cannot be fully validated. It requires a deterministic three-pass protocol and cannot bypass data, logic, or propagation checks.