You do not run each skill by hand. You set ONE scheduled task in Claude, and it runs a set of skills as parallel agents. Your inbox is handled before you sit down, and again after lunch.
Set once in Claude under Scheduled Tasks. It runs on its own from then on.
One instruction tells Claude which skills to run and the rule it can never break. Copy it into the scheduled task.
Process my inbox by running the following skills as parallel agents: invoice to draw, meeting prep, unsubscribe, and update CRM notes. Draft everything for my review. Never send or move money without my okay.
That one instruction spawns each skill as its own agent, all working your inbox at the same time. Right now only the first one is built. The rest are examples of what comes next.
Files and renames your Cedar Terrace invoices, enters the line items into the draw package, asks you before the category split. Never touches the money.
For your next call, pulls the last transcript and any open to-dos and drafts a one-page prep sheet.
Flags newsletters and clutter so your inbox is only the mail that needs you.
Spots a new investor in your email, drafts a contact for Go High Level on do-not-disturb, and asks you to approve adding them to the pipeline.
Each new skill you build with the agent builder just gets added to this one instruction.
Your email triaged, filed, and drafted before you sit down, and again after lunch. You review and approve. Nothing sends without you.