citation-guard

Verify citations and factual claims in AI-generated content against a research corpus.

Updated Mar 11, 2026
One-click install
npx skills add https://github.com/robit-man/transcribe-cli --skill citation-guard
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: citation-guard
Source: https://github.com/robit-man/transcribe-cli/tree/main/.claude/skills/citation-guard
Command: npx skills add https://github.com/robit-man/transcribe-cli --skill citation-guard

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill prevents AI-generated content from hallucinating citations or making unsubstantiated factual claims by automatically verifying references against a known corpus.

Core Features & Use Cases

  • Citation Interception: Detects and extracts citations (REF-XXX, DOIs, @-mentions) from generated content.
  • Corpus Verification: Checks if extracted references exist within specified project directories (.aiwg/research/sources/, .aiwg/research/findings/).
  • GRADE Compliance: Assesses if hedging language in claims matches the quality of the cited evidence.
  • Use Case: When an agent writes a report citing "REF-123", this skill checks if "REF-123" exists in the research directory and if the claim's strength aligns with the source's quality, blocking or warning if not.

Quick Start

Use the citation-guard skill to verify all citations in the draft report before it is finalized.

Frequently Asked Questions about citation-guard

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

FAQPage Schema
How do I stop AI from hallucinating citations in generated research reports?

To stop AI from hallucinating citations, you can automatically verify references against a known research corpus. This skill intercepts citation generation, checks extracted references against project directories, and blocks any unsubstantiated factual claims.

How do I verify if a DOI or REF tag exists in my project corpus?

You can verify if a DOI or REF tag exists by checking extracted citations against specified project directories like `.aiwg/research/sources/` and `.aiwg/research/findings/`. It validates the existence of references within these known corpus locations.

What is GRADE compliance for assessing hedging language in AI content?

GRADE compliance for hedging language assesses whether the strength of claims matches the quality of cited evidence. It validates if hedging language aligns with source quality, issuing warnings or blocking content for GRADE violations.

Can I use citation verification for legal compliance reports?

Yes, you can use citation verification for legal compliance reports. It ensures content generation complies with quality assessment standards by automatically checking references and preventing unsubstantiated claims from proceeding in legal documents.

What happens when an AI agent generates content with uncitable claims?

When an AI agent generates content with uncitable claims, the system blocks the hallucinated citations and tracks the uncitable claims for future reference. It prevents the unverified content from proceeding until references are properly substantiated.

Do I need a specific research directory structure to validate citations?

Yes, you need a specific research directory structure to validate citations. The skill checks references against designated project directories, specifically looking for files within `.aiwg/research/sources/` and `.aiwg/research/findings/` folders.