skills.rest
ultraviollettnympho
ultraviollettnymphoCOMMUNITY
Explore all skills by this author →
GitHub

doublecheck

Verifies AI-generated claims through web search and adversarial review with sourced reports.

1|1|Updated Jan 29, 2026
Search Tags:#fact-checking#verification#web search#hallucination detection#adversarial review#ai output#citation validation
One-click install
npx skills add https://github.com/ultraviollettnympho/transit-ticket --skill doublecheck-ultraviollettnympho
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: doublecheck
Source: https://github.com/ultraviollettnympho/transit-ticket/tree/main/.github/skills/doublecheck
Command: npx skills add https://github.com/ultraviollettnympho/transit-ticket --skill doublecheck-ultraviollettnympho

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) components.

What problem does it solve? AI-generated text often contains fabricated citations, invented statistics, and confident-sounding errors that are hard to spot manually. This Skill extracts every verifiable claim from AI output, finds independent sources, and flags hallucination patterns so users can check facts before relying on them. ## Core Features & Use Cases - Three-Layer Verification Pipeline: Extracts and categorizes claims (factual, statistical, citation, entity, causal, temporal), searches the web for supporting or contradicting sources, then runs an adversarial review targeting known hallucination patterns. - Active and One-Shot Modes: Run persistent inline verification on every substantive response, or verify a specific piece of text on demand with a full structured report. - Risk-Rated Reporting: Rates each claim as VERIFIED, PLAUSIBLE, UNVERIFIED, DISPUTED, or FABRICATION RISK, with source links and a structured report template. - Use Case: A lawyer receives AI-drafted analysis citing several cases. Running doublecheck reveals one case citation cannot be found in any legal database, flagging it as FABRICATION RISK before it reaches a filing. ## Quick Start Ask the assistant to doublecheck the previous response and produce a full verification report with sources.

Frequently Asked Questions about doublecheck

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

FAQPage Schema
How do I verify AI-generated text for hallucinations?▼

Invoke doublecheck with the text you want verified, or activate persistent mode to verify every substantive response. It extracts each verifiable claim, searches for sources, and rates claims from VERIFIED to FABRICATION RISK with links you can check.

How does doublecheck detect fabricated citations?▼

It searches for the exact case name, statute, paper title, or standard referenced in the text. If the citation cannot be found anywhere, it is flagged as FABRICATION RISK, since models frequently generate plausible-sounding references to documents that do not exist.

What is the difference between active mode and one-shot mode?▼

Active mode runs continuously, appending an inline verification section to each substantive response and escalating to a full report for high-risk findings. One-shot mode runs the complete three-layer pipeline once on a specific text you provide.

Can doublecheck verify legal and regulatory content?▼

Yes, and it treats legal content as high-stakes, always producing a full verification report rather than inline summaries. It gives extra scrutiny to case citations, statutory references, jurisdictional claims, and regulatory interpretations.

What are the limitations of automated fact verification?▼

Web search may miss recent or paywalled sources, and the adversarial review uses the same underlying model that may have produced the original output. A VERIFIED rating means a supporting source was found, not that the claim is definitely correct.

Related Skills

Explore More Content & Communication Skills→
shysssthanhtrishysssthanhtri

doublecheck

Extract verifiable claims from AI-generated text and verify them via web searches.

Community
Advanced
Kgan01Kgan01
6

hallucination-detector

Detect fabricated URLs, unsourced statistics, fake citations, and unfounded future claims in agent outputs.

Community
Advanced
sure-scalesure-scale
72

legal-research

Retrieves current statute and regulation text from official primary legal sources via webfetch.

Official
Intermediate
EliasOulkadiEliasOulkadi
110

research

Verify factual claims via web search with evidence hierarchy and source credibility checks.

Community
Advanced