The poster built an n8n workflow that sorts unread Gmail into leads, support, newsletters and other, labels each email, and has an AI model write a reply that is saved as a draft. Nothing is sent automatically. Every run showed as successful, but the drafts had problems: one offered a 15% discount nobody approved, all were signed "[Your Name]", the recipient field was empty, and one run made two drafts for the same customer. He fixed it with clear rules in the prompt, the right recipient setting, a "processed" label and marking emails as read. The workflow file is free on GitHub.
PromptWrite a short reply to the email below as [YOUR NAME] from [BUSINESS]. Rules: never offer discounts, refunds...
Looking back over many Hermes Agent sessions, the poster says one of his biggest lessons came from a slide deck where one slide said 14 and another said 10 for the same figure. Now he has the agent show how it got every number and double check them before he sees the result.
PromptBefore you give me the final version, list every number in it with where it came from (file, page or...
The poster's Grok Bot looks through the obstacle photos his robot vacuum takes every night. One morning it told him his dog Pretzel was due for a groom: in the clearest photo, the hair over her eyes was in her face and her leg fur was long, so it said to do the eyes first.
PromptEach morning, look at the newest photos in [WHERE THE PHOTOS ARE SAVED]. If you see my dog [NAME], tell me if...
The poster runs several Grok Bots and shared the first message he gives each new one: a name, a role, one job and a list of things it never does. Every bot has to ask him before anything involving money, signatures or posting in public, and write down how it was set up the same day.
PromptYou're my [ROLE], called [NAME]. Your only job is [ONE JOB]. Never [THINGS IT MUST NOT DO]. Before anything...
Axios reported on Oct 22 that big US banks are finding AI agents inside customer accounts, often signed in with the customer's own password. JPMorgan, Bank of America, Citi and Morgan Stanley are restricting agent access, and a Chase spokesperson said customers shouldn't give their username and password to anyone. Most banks are moving toward read-only access instead.
PromptHelp me with my bank account without logging in for me. I'll download my last [3] months of statements as PDF...
The poster's agent reminded him that his home and contents insurance was about to renew at a higher price. He asked Grok Bot to find alternatives, and it came back with a better quote. He then connected a voice tool so the agent could call his current insurer, and says that once they shared the lower quotes, the insurer cut his renewal price. He says it took about 10 minutes in total.
PromptMy home insurance with [INSURER] renews on [DATE] at [NEW PRICE] a year, up from [OLD PRICE]. Here is my...
The poster had Grok Bot write a quick email to his hotel asking whether they could offer a better rate. He says the front desk called him back and lowered the price from $330 to $240 a night, and that the whole thing took about 30 seconds of his time. His point: most people never ask.
PromptI have a booking at [HOTEL NAME] from [CHECK-IN DATE] to [CHECK-OUT DATE] at [CURRENT RATE] a night,...
In an interview posted by Marina Mogilko, Felix Rieseberg, who helped build Claude Cowork at Anthropic, explained how he catches confident mistakes. He has Claude start five threads that each try to prove a document wrong, then a separate Claude judges which points hold up. Every number gets checked against a primary source, and he rarely lets a session grade its own work.
PromptHere is my draft: [PASTE OR ATTACH]. Start 5 separate reviewers, each with one job: find factual errors, find...
The poster says keeping up with the admin for his rental, like bank transactions and utility invoices, used to need a coding agent and screen-scraping scripts. Now he says ChatGPT Work handles much of it, including running it on a schedule.
PromptOn the [DAY] of every month, go through the utility invoices in [FOLDER OR EMAIL LABEL] and the bank export I...
The poster says ChatGPT kept defaulting to Work mode, even for simple questions and searches, which used up his Codex quota. Work and Codex draw from the same allowance, so quick questions asked in Work eat into the limit you may want for longer jobs.
PromptBefore you start a Work task, tell me in one line what you plan to do and roughly how many steps it takes. If...