Engineer pipeline with stage shares

The idea

Work follows an engineer's pipeline, and each stage gets a share of the time and cost. Review is one plan review and one holistic diff review, within about 10% of spend.

The problem
Agents spent their budget on research and review loops instead of building.
Why it matters
Most of the build time goes into writing and testing code.
Where it stands
In production as of

A step-by-step replay of this idea is not published yet. The library adds one when a recorded run shows it.

  • One effort allocation per item and stage

    One decision per item and stage picks low, medium or high effort. That level sets the stage's turns, time, cost, rounds and consultations, and the worker can ask for more.

  • Team pipeline profile and layered preferences

    Onboarding proposes how the team works (design first, test first, trunk, approvals, strict paths, stage shares), with a backtest over 30 merged PRs. Accepted settings resolve in layers, with the person's own layer on top.

  • Tier × effort routing from runtime signals

    Each call's model tier and effort are chosen separately. The inputs are the stage, the risk of the pending action, the change size, open findings, failures, repository size and the budget left. Floors tune from verified outcomes.

  • Action fusion: edit plus check

    An edit can carry its planned check. The runtime runs the check as its own gated call in the same model turn.

Turn the numbers into shipped work.

Crewshal runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.