Code checks for code that AI workers operate

The idea

A check without a model holds worker code to module size, header, import cycles, colocated tests, index entries and typed exports. It runs at act exit, in verification and in the platform verifier.

The problem
Code that agents write can become hard for later agents to find and change.
Why it matters
The code stays easy for LLM agents to read and operate.
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.

  • Opinionated harness per work type

    Each work template carries steps, a quality bar, pitfalls, tool guidance and practitioner notes, plus a minimum model tier and effort. Its named industry checklists are settled by detectors or by attestations with evidence.

  • Stage exit checklists and loop limits

    Each stage has an exit checklist. Code judges the clear items and System One the unclear ones. Repeated holds go to a person, and a loop guard stops the run when the same result repeats.

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.