Agents can do anything.They still need you to steer

Twelve AI sessions stream walls of text. One ends with four decisions whose context is buried two screens up; answering them carefully takes so long that more sessions keep asking, and the queue never ends. Let's empower you to make the most of it: with crafters, each question lands on the work it is about, with its context: a product page, a checkout, a slide, an email, a code diff. You decide in place, collect your decisions, and send them all at once; every agent goes back to work.

Pick your job, say founder. Its three crafts sit side by side in crafters, each with the agent driving it, its sub-agents, the tools it's made with and the ones it's connected to. Open one: the work fills the window, and the agent's question lands right on it, where you answer it.

Encode it: a skill gains an engine that renders the work, roles and tools, and gates where the agent stops for your call; its process.yml grows alongside. Your team adopts it: shared processes, each with its version, its gates and who uses it; try a new version first or roll it out to everyone. Anyone can adopt it: seven engines, surface, data, logic, structure, document, timeline and model, each running processes people publish, free or paid.

Your job's work asks one question on a point, say whether to backfill three missing emails. You answer: leave them null. crafters reads the pattern under your call, never invent missing data, and the next run of that process, the checks the platform writes, and all future work follow it.