Ask your agent for real things.
Every prompt below runs on your own accounts through Nilyo, in ChatGPT, Claude or any MCP-compatible agent. Copy one, replace the names, and see what your agent answers.
Search & research
Find the right person or company on your own LinkedIn, then keep working with exact profiles instead of guesses.
“Who do I know at Stripe on LinkedIn?”
Lists your first-degree connections at the company, with titles, so you can pick the warmest path.
“Find five heads of sales at fintech companies in Paris.”
One precise people search with location, title and industry filters; results come back with stable profile IDs.
“Open https://www.linkedin.com/in/… and tell me what this person does, then check if we ever emailed.”
Resolves the profile from its URL, then searches your mailbox for the same person before you reply.
“What does this company do and how many employees does it have on LinkedIn?”
Company lookup with headcount, industry and website, ready for a meeting brief.
“Look at this Instagram creator: followers, bio and recent activity.”
Profile research on your own Instagram account, without a scraper.
“Is +33 6 12 34 56 78 on WhatsApp?”
Checks the number before you start a conversation.
Inbox & summaries
Catch up across every inbox in one answer, with the counts and dates taken from the mailbox itself.
“What came in on LinkedIn, WhatsApp and email since yesterday? Summarize it.”
Reads the recent conversations of each account and groups them by channel, newest first.
“Which conversations still need an answer from me?”
Flags threads where the last message is not yours, with a proposed reply for each.
“How many emails did I receive today in my OVH mailbox?”
Uses the folder counts of the mailbox, not a guess from one page of results.
“Find the last email from Sarah with an attachment and show me the file.”
Searches the mailbox, opens the message and returns the attachment (images are displayed inline).
“Summarize the WhatsApp group “Product launch” for the last week.”
Reads the group history and gives you decisions, open questions and who said what.
“What did Paul answer on Telegram about the invoice?”
Finds the chat by name and quotes the actual message.
Reply & send
Say what you want sent. Nilyo resolves the person, shows the draft when needed and sends from your own account.
“Send a WhatsApp to Julien saying I'll be 10 minutes late.”
Sends immediately when exactly one Julien matches; asks which one otherwise.
“Reply to the last LinkedIn message from Marie: yes for Thursday 3 pm.”
Reads the conversation first so the reply fits the thread.
“Answer this email in a friendly tone and cc my colleague.”
A true reply (same thread) with the recipients you asked for.
“Send a voice note to Nico on WhatsApp with this text.”
Native voice message, playable in WhatsApp.
“Draft a reply to every unanswered Instagram DM and let me approve them one by one.”
Drafts first, sends only after your approval, one at a time.
“Send a connection request to this profile with a short note about our call.”
Invitation with a note under 300 characters, from your own LinkedIn.
Network & content
Grow and animate your LinkedIn presence with a reviewed batch every day, never bulk automation.
“Show me the latest posts from my top prospects and draft a comment for each.”
Fetches each person's latest post, proposes one comment per post, comments after approval.
“Who reacted to my last post? Anyone I should follow up with?”
Lists reactions and comments, highlights people in your target accounts.
“Write a LinkedIn post about our new feature and publish it when I say go.”
Draft, iterate, then publish from your account or a company page you manage.
“List the new applicants on my job postings and summarize the three strongest.”
Applicants and resumes from your LinkedIn job postings or Recruiter projects.
“Who follows me on Instagram that I don't follow back?”
Followers and following lists from your own account.
Calendar & meetings
Prepare and confirm meetings with the context from your accounts.
“What's on my calendar tomorrow, and who are these people on LinkedIn?”
Events with attendees, then a profile lookup for each external attendee.
“Create a 30-minute call with Sarah on Friday morning and confirm by email.”
Creates the event, then sends the confirmation from your mailbox.
“Before my 2 pm meeting, brief me: LinkedIn profile, our last emails, our last WhatsApp messages.”
One brief built from four accounts, nothing invented.
Cross-channel
The reason to connect several accounts: context from one channel, action on another.
“Check what we said by email before I answer this LinkedIn message.”
Mail history first, then the LinkedIn reply with the right context.
“When this lead answers on LinkedIn, send me a WhatsApp.”
A persistent agent (OpenClaw, Hermes, n8n) receives the LinkedIn event and notifies you on WhatsApp.
“Find this person on LinkedIn, then follow up on WhatsApp if we have their number.”
Profile lookup, contact match, then a message you approve.
“Every morning at 8, check email, LinkedIn and WhatsApp and send me the digest.”
Scheduled task in ChatGPT, or a persistent agent that runs on its own.
What happens when you send one of these
Resolve first
Names and URLs become exact profile, chat or email IDs before anything is done. Two matches: the agent asks.
Show before sending
Sends, posts and invitations are separate actions; the agent shows the text unless you already gave it word for word.
Paced like a person
Daily budgets per account (about 100 LinkedIn actions, 20 new WhatsApp chats) and one call at a time.