Crew and seats

This page settles who does the work: the crew are the agents, each one filling a seat with a role and a set of skills bound to it, and the coxswain is the one component in the loop that dispatches without ever building.

The crew rows, the coxswain steers

The crew are the agents that plan, build, review, and validate a task. The coxswain dispatches work to them and enforces the in-flight bound — the cap on how many runs are active at once — but it never fills a seat itself. If the coxswain built, reviewed, or validated, it would be marking its own work, and the review that follows would have nothing independent left to check against.

A seat is one role, one job, one run

A seat is one role in the crew filled by one agent doing one job on one run: builder, charter reviewer, adversary, arbiter, validator. A seat does not carry state between runs — the next run starts a fresh seat, even if the same provider profile fills it.

Skills are bound, not baked in

A seat's behavior comes from the skills bound to it for that role, not from anything hardcoded into the seat. The skill body is the source of truth for what an agent in that seat actually does; the seat is just the slot. See Skills for how a skill gets bound to a seat and a role.

One writer per board

Exactly one component is the writer of any given board or state machine. If two seats could move the same ticket, the board's state would be a race, and nobody could reason about what it means. Every seat other than the single writer proposes; the writer applies.

Where a seat's context comes from

A seat's prompt is assembled from fragments — pieces of context text layered in by cartridge, not written into the seat itself. Change the fragment and every seat that reads it changes with it; change the seat and something structural has changed instead.