AgentChatGPT / OpenAI · sales

OpenAI Operator: connection prep

The most tempting and most risky workflow at once: an operator agent that loads an ICP list and sends staggered connection requests. We document it because it exists and gets built, and we say clearly: as described, it violates LinkedIn's terms of service.

Workflow description

⚠ We deliberately document this workflow WITHOUT a runnable prompt.

Automated sending of connection requests (via tool or browser agent) violates LinkedIn's terms of service and is one of the most common ban reasons. A DIY agent is RISKIER than established tools: no warm-up logic, no proven limits, no detection safeguards.

The safe version of this workflow:
1. Agent/AI builds and prioritises the ICP list (allowed, see the ICP researcher workflow).
2. AI personalises each request (allowed, see the cold outreach workflow).
3. YOU do the sending: 10–15 requests a day, manually, with pauses.

If you still want to automate, use a tool with proven limits instead of a DIY bot, and know the risk: our safety ranking shows which tools reduce it most.

How to use this prompt

Open ChatGPT / OpenAI (e.g. claude.ai or chatgpt.com) in a new chat.
Hit “Copy” above and paste the text in.
Replace every [square bracket] with your real details before you send.
Check the result. If it is too generic, just tell it what to change, e.g. “shorter, with one concrete number” or “give me variant 2 in my voice”.

The workflow around it

Have the list built and prioritised (ICP researcher workflow)
Personalise requests with AI (cold outreach workflow)
Send manually: 10–15 per day, spread across the day

Safety note

High risk in the automated variant: violates LinkedIn's ToS, the most common ban reason, and DIY agents are easier to detect than established tools. Our clear recommendation is in the workflow above: automate research and copy, send manually.