GET YOUR FREE
Personal AI Agent- Connect Checklist
Check what an AI agent can read, reach, install and undo
before you connect it to a real account.
One PDF for anyone about to hand an agent their email, calendar or browser who wants the four checks, the settings and the sources on one sheet.
You don’t need to guess what a new agent can touch anymore. This checklist is built for one thing: running a short safety check before any AI agent gets access to a real account.
What’s Inside:
– What Will It Read?: Four tick boxes that keep hidden instructions in web pages and emails away from your sensitive accounts.
– What Can It Reach?: An account table to fill in, plus the least-privilege habits that keep a fooled agent boxed in.
– What Did You Install?: Five checks for any skill, plugin or extension, built on the pattern behind 341 malicious ClawHub skills.
– What Can’t Be Undone?: A receipt-style approval check and a table of which actions stay on manual approval.
– Where the Switches Are: The approval and security settings for Claude in Chrome, ChatGPT Work and OpenClaw, with a space to log your own setup.
Every page uses published data from sources like Anthropic’s and OpenAI’s own safety pages, Koi Security, Palo Alto Networks Unit 42 and OWASP, so you understand the strategy, not just the shortcut.
Still connecting agents to accounts and hoping for the best?
This Is For You If:
– You have 5–20 hours a week max and need leverage, not more grind.
– You’re setting up an AI agent to handle email, calendar or browser tasks and want it working without handing over every account.
– You want to use Claude in Chrome, ChatGPT Work or OpenClaw, but don’t know yet which settings actually limit what they can do.
– You’re tired of clicking “approve” without reading, and want a quick check that tells you which actions really need a look.
The problem isn’t your ideas. It’s that you don’t have a system to check an agent before it gets access. This checklist is your system.
This free resource is built for busy professionals: people who want AI systems that give hours back, around a full-time job and a family.