skills.rest
williamsforeal
williamsforealCOMMUNITY
Explore all skills by this author →
GitHub

hallucination-protocol

Flag unverified factual claims with [VERIFY] tags in outputs.

1|Updated Feb 6, 2026
Search Tags:#verification#marketing analytics#guardrails#anti-hallucination#fact checking#ad copy compliance#uncertainty tagging
One-click install
npx skills add https://github.com/williamsforeal/Cyclone-SS --skill hallucination-protocol
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: hallucination-protocol
Source: https://github.com/williamsforeal/Cyclone-SS/tree/main/.cursor/skills/hallucination-protocol
Command: npx skills add https://github.com/williamsforeal/Cyclone-SS --skill hallucination-protocol

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents the AI from presenting unverified facts as certainty by forcing explicit verification whenever outputs involve numbers, dates, platform behavior, pricing, or other “current” or factual claims.

Core Features & Use Cases

  • Verification-first guardrails: Flags any statement that depends on information not provided in-context, including metrics, spend, ROAS, CAC, prices, deadlines, and “typically/usually” assertions.
  • Confidence labeling and auditability: Requires inline [VERIFY] tags or a research-draft list when verification is missing, so deliverables are not produced with fabricated certainty.
  • Operational triggers for complex operations: Governs ad spend decisions, copy with factual claims, platform policy assertions, competitor metric references, and any third-party study/testimonial/stat citations.
  • Escape hatches for missing data: Refuses to invent numbers, instead offering computation frameworks, assumption-based ranges, or requesting the exact inputs needed to compute results.

Quick Start

Ask the AI to produce or advise on your Meta or Shopify marketing plan, and require it to mark every factual/numbered claim with [VERIFY] unless you supply the source data.

Frequently Asked Questions about hallucination-protocol

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

FAQPage Schema
How do I stop AI from fabricating marketing metrics and unverified claims?▼

To stop AI from fabricating marketing metrics, use anti-hallucination guardrails that flag unverified statements with [VERIFY] tags whenever outputs involve ROAS, CAC, prices, or platform behavior assertions.

How do I validate ad copy compliance for factual claims and platform policies?▼

You validate ad copy compliance by applying operational triggers that require sources for competitor metric references and platform policy assertions. The AI must flag any factual claim lacking in-context source data for human review.

What happens when AI cannot verify spend recommendations or factual numbers?▼

When AI cannot verify spend recommendations or factual numbers, it refuses to invent data and instead offers computation frameworks, assumption-based ranges, or requests the exact inputs needed to compute results.

Does this anti-hallucination approach work for Meta and Shopify marketing plans?▼

Yes, this anti-hallucination approach works for Meta and Shopify marketing plans by marking every factual or numbered claim with [VERIFY] unless source data is provided, ensuring operational decisions rely on auditable facts.

Why does the AI output [VERIFY] tags instead of direct answers?▼

The AI outputs [VERIFY] tags instead of direct answers because confidence labeling enforces auditability. It flags statements depending on information not provided in-context, preventing deliverables from being produced with fabricated certainty.

Related Skills

Explore More Legal & Compliance Skills→
ronnycodingronnycoding
20

secure-web-search

Perform privacy-preserving web searches with source verification and fact-checking.

Community
Basic
JDMAXiliusJDMAXilius

social-media-analyzer

Calculate engagement, click-through, and return on ad spend from social campaign data.

Community
Intermediate
vivy-yivivy-yi
363

roi-analysis

Calculate marketing ROI and ROAS for Xiaohongshu campaigns.

Community
Advanced
MDGrey33MDGrey33
2

research

Coordinate multi-depth research workflows that gather evidence and validate claims with citations.

Community
Advanced