Tasks live in two lists, and the split is the point.

Why not one list

"Get a Stripe API key" and "the pricing page overflows on mobile" are both tasks, and they are not the same kind of thing. One needs a person with a card and a browser. The other needs an agent with the repository.

Put them in one list and the agent ticks the first as done without doing it. That is worse than leaving it open, because a false tick stops anyone looking again — and you find out weeks later that the key was never created.

So the split is in the data, not in a label. A task has an owner: you, or the agent.

What belongs to a person

Anything requiring a card, an account, a decision, a legal choice, or a conversation. Buying a domain. Approving copy. Choosing between two designs.

What belongs to an agent

Anything doable from the repository. A failing test. A layout bug. A migration. A refactor.

Moving one across

Use the move control on the row. Tasks that turn out to be misfiled are common and moving them is cheap — what matters is that the two lists never merge.

Where they come from

You can add them by hand, and they also arrive from the browser extension: a capture with a comment becomes a task against the site it was taken on.