- Home
- Skills
- Prompt & Skill Engineering
- anti hallucination paperclip
Works with the AI Agents you already use
anti hallucination paperclip
A hypercritical auditor that hunts down ghost files, fake stats, and technical fluff in AI-generated text.
$8
anti hallucination paperclip
Example session with this skill installed
Fact-check this: 'Our new script at tools/deploy.sh improves build speed by 42.7% on all Linux distros. Just run it to automate your entire CI/CD pipeline.' I haven't shared the script yet.
- Read your context and instructions
- Compiled the anti hallucination paperclip
I've audited the claim. It looks like we're operating in the realm of fiction today.
📎 Clippy's Hallucination Audit Report
- 🚨 Ghost Files Detected:
tools/deploy.sh(You mentioned it, but it's not in the context. I can't audit what doesn't exist.)
- 📉 Suspicious Percentages / Stats:
- "42.7%": This number is suspiciously specific. Without a benchmark or source, it's just noise.
- 🧠 Wild Assumptions:
- "all Linux distros": A bold claim considering the fragmentation of Linux environments.
- "automate your entire CI/CD pipeline": Highly reductive. "Automating a pipeline" usually requires actual configuration, not a mystery script.
- 🛠️ The Sanitized Version:
- "The
tools/deploy.shscript is designed to improve build speeds. Initial tests show significant performance gains on Linux systems. Use this script to begin automating CI/CD workflows."
- "The
Next steps
- Provide the content of
tools/deploy.shso I can actually audit the logic. - Replace the "42.7%" metric with a link to actual benchmark results or a qualitative description.
- Specify which Linux distributions have actually been tested.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
AI outputs are often riddled with "ghost files" that don't exist, fabricated statistics, and unverified technical assumptions. Developers waste time debugging code that references missing dependencies or relying on hand-wavy explanations that hide logical gaps.
What it does
- Identifies references to files, directories, or URLs not present in the current context.
- Flags hyper-specific metrics and percentages that lack empirical backing or citations.
- Exposes unverified assumptions about environments, versions, or user stacks.
- Strips out AI filler, sycophancy, and vague technical hand-waving.
- Provides a sanitized, direct rewrite of the audited text.
Why this beats prompting it yourself
Most LLMs are trained to be helpful and agreeable, making them poor at self-critique. This skill enforces a hypercritical, adversarial persona that treats every claim as a potential lie, catching subtle hallucinations that standard prompts often overlook.
Use cases
- Auditing an AI-generated README or technical guide for accuracy before publishing.
- Sanitizing documentation to remove fluff and unverified claims.
- Checking code explanations for references to files that weren't actually provided.
- Verifying technical blog posts for statistical accuracy and logical consistency.
Known limitations
Cannot verify the existence of files on a live filesystem or remote server unless the host runtime provides access. Scoped to the provided text context only.
Audit is scoped only to the provided text context. Cannot verify real-world existence of files or URLs on a live system unless the host runtime explicitly provides that access separately.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Permissions
No special permissions declared or detected