# Steal this setup: a Verify lane

A standing chat, a queue of your agent's uncertain claims, evidence first, one question per message, and no all-clear while the data is not clean.

From the [Templates & checklists](../index.md#templates) section of [The Muse Field Guide](../index.md).

## What it is

A standing practice, run as its own chat, that audits your agent's stored claims with you as the court of last resort, one claim at a time.

## Parts list

- A queue of uncertain claims, drawn from memory, records, or a graph underneath.
- The house rules: one question per message; evidence first; never "nothing to do" while data is unclean.
- Governed writes: confirmed edits allowed, merges stay your decision, speculation held.
- A status habit: every update names the next batch in motion.

## Steps

1. Collect the claims nobody has confirmed into a queue.
2. Exhaust the agent's own sources for each claim before asking.
3. Ask one question, with evidence and a single tap option.
4. Write the ruling back with provenance, and correct the record in place.

## Where this came from

The rule the lane enforces predates the Muse seat and was bought by a teardown: an earlier knowledge graph was rebuilt under one rule, no evidence, no edge. The lane is that rule, still being enforced one question at a time.

Deep dive: [Memory that holds](../deep-dives/memory.md)

Related resources: [The Verify lane: one claim at a time](the-verify-lane.md), [verify-loop: audit memory one claim at a time](verify-loop.md)

Canonical page: https://www.niclydon.io/muse/items/verify-lane-starter.html
