Security · The record
An action needs a record that survives the agent.
Draft for review. This page describes the proposed design. Features without a confirmed implementation status are marked for review; they are not claims of running controls.
An agent uses a model to reason and a harness to act. The harness is the agent program that calls a model and acts on the answers with its own tools. A seat is the confined environment in which the harness runs. The Awakener is the mechanism being built to start an agent within a limit and record what happened. A receipt records who asked, which agent ran and how it ended. A ledger is the service’s ordered record of events.
A conversation tells part of the story. The security design also needs a record of permission, action and outcome that the acting agent cannot rewrite.
This draft describes the record’s role. It does not claim that receipts for complete live Awakener runs have been demonstrated.
Design outline dated 6 October 2026. What the four status labels mean.
The proposed design
Each heading states the design or practice. Its status shows whether it is in use, built, in build or designed; a built part does not establish that the whole control is running. Where the outline gives no implementation status, that status awaits review.
A receipt for every run.
The proposed record ties a run to the work it was permitted to do and records how it ended. A receipt is the basis for checking what happened after the agent has stopped.
Implementation status awaits review.
The record has its own service.
The outline places the ledger behind a service boundary. Agents should not be able to edit its stored history directly; the record is kept separately from their working text.
Implementation status awaits review.
Changes are appended, versioned and timed.
The design calls for changes to add a new event rather than erase the old one. Versions distinguish successive states, and times put the changes in order.
Implementation status awaits review.
Usage belongs to a run.
The design associates usage with the run that incurred it, so limits and outcomes can be assessed together. This does not claim that a run can be stopped on a token count; tokens are the units of text processed by a model.
- Designed
What this provides
The intended result is a record that can be checked after an agent stops, without relying on that agent’s account of its own work.
What is not yet proven
The complete path from permission through a live Awakener run to a receipt remains unexercised. Reviewed parts of the record are not proof of that complete path.
The service boundary, history rules and retention of receipts need confirmed implementation statuses and evidence before this draft can claim they are running.