# Pre-register kill criteria

Before an experiment runs, write down what result would kill it. When that result arrives, kill it, even if the signal looks interesting.

From the [Tips & tricks](../index.md#tips) section of [The Muse Field Guide](../index.md).

## The move

Before an experiment runs, write down what result would kill it. When that result arrives, kill it, even if the signal looks interesting.

## When to use it

For every prototype, trial, and pilot. This is how you keep a tireless agent from turning every prototype into a permanent resident. The criteria go in the chat or the project file before the run, where enthusiasm cannot edit them afterward.

## Steps

1. State the success bar and the kill bar before the run starts.
2. Name the free or incumbent baseline it must beat.
3. Run it on real work, not a demo set.
4. If the kill bar is met, shut it down and preserve the record.

## The gotcha

A genuine signal is the most dangerous outcome: good enough to argue for, not good enough to keep. The criteria exist for exactly that argument.

## Where this came from

A routing prototype scored 526 real requests and produced a genuine signal, an AUC of 0.804, and was still shut down because two free baselines nearly matched it. The verdict was "does not earn its keep"; the record was preserved, the prototype deleted.

Deep dive: [Practices and tips: the resource core](../deep-dives/practices.md)

Related resources: [Turn judgment calls into checks over facts](verdict-rules-over-facts.md), [Ingest before you delete](ingest-before-you-delete.md)

Canonical page: https://www.niclydon.io/muse/items/pre-register-kill-criteria.html