What does your agent do when it needs a human?

leafslip launches here on Thursday, 24 September.

The short version: your agent writes a form as a plain-text file, sends it to the person who can answer, and picks the answer back up as a real true or false it can act on. If nobody answers in time, the request expires with no answer at all. Silence never counts as a yes.

Before the launch, I'd love to hear from anyone running agents on real work:

What does your agent get stuck on? A permission it doesn't have, a missing credential, a judgement call, something else?

How does it reach you today? A Slack ping, a log line nobody reads, or does it just guess?

Who has to answer? Someone on your team, or someone outside it, like a customer or a reviewer at another company?

I'll read every reply, and the answers will shape what leafslip builds next.