Completion truth gates

The idea

New tests must run here or have a trace for each assertion. Claimed behaviour must be tested, and the same bug elsewhere fixed or ruled out. Planned changes must exist, steps tick only on evidence, and spend over budget needs a recorded decision.

The problem
An evaluator's 'met' stood in for unrun tests, fixes missed siblings, and plans quietly went unfinished or over budget.
Why it matters
Execution and evidence decide done, and the receipt shows which planned steps and costs held.
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.

  • Acceptance criteria as executable checks

    At plan time, each acceptance criterion maps to a test, script, query, HTTP check or screenshot check. Verification runs all of them in each round.

  • Revert proof

    A new or changed test must fail when the fix is reverted.

  • Plan contract and plan rigor

    A plan must map criteria to checks, close each research finding with a step or an answer, and base its claims on code that was read. The gate corrects the plan instead of failing it.

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.