6 November 2026

Why I built SuspectFraud after a suspicious recruitment message

A suspicious recruitment message led me to build an assistant skill for checking warning signs, preserving evidence and preparing reports.

Why I built SuspectFraud after a suspicious recruitment message

A WhatsApp account with a blue tick sent me a recruitment pitch. The badge looked reassuring, but the offer felt too good to be true. I wanted to check whether the person actually represented the company they named.

That distinction mattered. An account badge wasn't an answer to my recruitment question. I needed to check the claim through the company's own channels.

Working through the messages, evidence and reporting steps led me to create SuspectFraud, an AI assistant skill that helps you assess suspicious content and organise a report.

Start with the message in front of you

Share a message, screenshot or link in a compatible assistant conversation and ask, “Does this look like a scam?” SuspectFraud instructs the assistant to give a qualified assessment before deeper research, explain the warning signs it can actually observe and suggest one practical next step.

For example, a request to pay a deposit to release promised earnings is a concrete reason for concern. A bare link with no context needs more information. The assistant should explain that difference instead of treating every unfamiliar message as fraud.

A warning is a reason to investigate. It doesn't establish guilt, authenticate a sender or guarantee that a message is safe.

Make the report useful

The skill helps preserve the original evidence privately and prepare redacted copies for reporting. Passwords, one-time codes, recovery phrases and unrelated personal details should stay out of the material you share.

It then helps find the appropriate official destination: a domain registrar, a relevant hosting or service provider, the company being impersonated and, for Singapore cases, ScamShield. Those destinations have different roles. A domain's nameservers alone don't tell you who hosts the underlying website.

External reports require your authority and suitable account or tool access. A drafted complaint and a confirmed submission are different outcomes; the assistant should tell you which one happened and retain any receipt.

What is available now

Version 0.2.0 and the source are public on GitHub. The ChatGPT plugin is saved privately in my account while its public directory listing is pending.

SuspectFraud is a set of assistant instructions. It doesn't install an inbox listener, background monitoring or automatic Dot notifications. A compatible host needs to make the relevant content available to the assistant. The skill can miss scams or flag legitimate messages, so review its reasoning before acting.

I wanted a repeatable way to move from “this feels wrong” to a clearer assessment and a useful report. That's what this first release is designed to help with.

See SuspectFraud, its support details and current availability.