Grok BotUndisclosed

Pre-payment due diligence: Grok Bot finds one phone used by multiple supposedly independent businesses — $14,000 deposit potentially saved

broke boyX (Twitter) · 2026-11-09

Read original

WHAT THE F*CK. Grok Bot just helped me expose a $14,000 scam. I was about to send a deposit to a contractor. Everything looked legitimate. Professional website. Verified-looking reviews. Real business address. Something felt off. I gave Grok Bot the company details and asked one question: "Find everything that doesn't add up." It started digging: 1)Compared company registration records. 2)Checked domain registration history. 3)Analyzed customer reviews. 4)Found inconsistencies in the business address. 5)Connected several supposedly unrelated complaints. Then it found something I completely missed. The same phone number was being used by multiple supposedly independent businesses. One afternoon of research potentially saved me $14,000. I didn't hire an investigator. I gave Grok a mission. AI isn't just answering questions anymore. It's investigating problems. Prompt + how to turn this into a Grok Bot skill below ↓

Business model

The author was about to send a $14,000 deposit to a contractor that looked legitimate (professional website, verified-looking reviews, real address). He gave Grok Bot the company details and a single question: “Find everything that doesn’t add up.” The bot compared company registration records, checked domain registration history, analyzed customer reviews, found address inconsistencies, and connected several supposedly unrelated complaints. It then surfaced that the same phone number was being used by multiple supposedly independent businesses. One afternoon of agent-assisted research potentially saved the deposit without hiring an investigator. A screen recording attached to the post shows the bot’s findings, including four businesses sharing one phone number. Readers can reuse the “single clear mission + multi-source cross-check” pattern before any large payment: let the agent surface contradictions, then decide yourself.

Workflow

  1. Give Grok Bot the single instruction to find everything that doesn’t add up
  2. Bot compares registration records with domain history
  3. Analyze review patterns and link apparently unrelated complaints
  4. Check address consistency and cross-reference phone number usage
  5. Author manually reviews and decides whether to send the deposit

Adaptation plan

Results and numbers are as reported by the original post author. Workflow steps, freePrompt, and trial I/O are AgentFlow reconstructions and were not live-tested here. $14,000 is potential avoided loss, not confirmed cash recovered. The full prompt text promised in the post was not found in the visible replies. The screen recording names Crestview Home Builders and three other businesses; the phone number itself was not legible. No verifiable public-record links provided.

Prompt

AgentFlow draft · Untested.

You are a due-diligence assistant. The user will provide the target company’s name, website, phone, address, and the amount they intend to pay. Your only task is to find everything that doesn’t add up. Check in order: 1) company registration records and incorporation date; 2) domain registration history and registrant; 3) anomalous patterns in customer reviews; 4) consistency of address and contact details; 5) whether the phone number is used by other supposedly independent businesses. Output format: list findings by check item, mark each as “verified / needs further human confirmation”, then give an overall risk summary. If key information (e.g., full legal name) is missing, ask the user first — do not guess. You must not send messages, make payments, or contact third parties on the user’s behalf. All conclusions are for the user’s reference only; the final decision remains with the user. If sources conflict, prioritize publicly verifiable official records and cite them. Untested.

Running costs and pitfalls

Results and numbers are as reported by the original post author. Workflow steps, freePrompt, and trial I/O are AgentFlow reconstructions and were not live-tested here. $14,000 is potential avoided loss, not confirmed cash recovered. The full prompt text promised in the post was not found in the visible replies. The screen recording names Crestview Home Builders and three other businesses; the phone number itself was not legible. No verifiable public-record links provided. No independent cost benchmark is available. Check subscriptions, permissions and external actions before running a workflow.

Author-reported results are not independently audited.

Original text and media remain attributed to their authors.