Recurring codebase audits and import

What it is

The audit half.

• A recurring per-project audit. A run is subscribed to a project and happens overnight rather than in your working hours, which is what makes it safe: the agent works while you are away and you read the result later. See Nighty Tidy (the full-vision audit panel).
• How a run actually works, and its limits. The sessions a run starts are tagged as its own, so it can tell its work apart from yours — and it refuses to touch what it did not create. The machinery, the boundaries and how to drive one from an operator session are in Nighty Tidy — how it works and its limits (part 2).
• Your own audit definitions. An audit does not have to be the built-in one; you can author the definition a run works to.
• Findings, kept. Each run produces findings and keeps them as reports with a run history, so a problem found once can be checked against a later run.
• Run one now. An audit can also be started on demand, and because that starts a real paid session it is approval-gated.
• The bundled audit skills. Under a run sit the eight bundled skills that do the compaction-resilient audit and turn findings into commits — the execution layer, documented in Audit Framework (bundled audit skills).

The import half.

• Bringing an OpenClaw setup in. Paste whatever address or token you already have and the connection fields fill themselves in; you can then have a Claude session recreate your scheduled jobs and agent personas here. Nothing ported ever runs on its own, and secrets are never carried across. See Port your OpenClaw setup into Omniscio.

Where to find it

• The audit panel is the Nighty Tidy panel — that is where a run is started and where its findings are read.
• The import lives in Settings → OpenClaw.

How it behaves

• An audit owns what it created. A run tidies and changes only the things it made; the rest of your work is left alone, which is what makes an unattended overnight run acceptable.
• Findings are records, not actions. A run reports; applying a finding is a separate, deliberate step.
• Starting a run costs money. An on-demand audit spawns a paid session, so it goes through approval — the scheduled path is what runs without you watching.
• The import is one-way and inert. It reads a setup and recreates it here; it does not keep talking to the old gateway and does not bring credentials with it.

Related

• Nighty Tidy (the full-vision audit panel) — the panel and its owned-vs-shared boundary.
• Nighty Tidy — how it works and its limits (part 2) — the machinery, the tagged sessions and the limits.
• Audit Framework (bundled audit skills) — the skills a run executes.
• Port your OpenClaw setup into Omniscio — imports a setup's schedules and personas.