This starter has no authentication. Build your private copy and verify access before adding org information.

Owner: [name]
Status: Proposal — not approved
Last reviewed: Not yet reviewed
Next review: [date]
Sources: [links or document versions]

AI boundaries

Work or action	May do independently	Requires human review	Reserved for people / prohibited	Accountable owner
[pilot activity]	[specific scope]	[review before which action]	[explicit boundary]	[name]

Examples

[Give concrete allowed and disallowed examples. A research-only editorial pilot could allow source briefs while reserving all publishable editorial copy for humans. A relationship owner may require personal editing and approval before any warm outreach is sent.]

Enforcement

See access-plan.html for configured controls and unverified gaps. These instructions alone do not enforce permissions.

Room to experiment

[Specify which people/teams may improve workflows locally, the limits of an experiment, and what requires coordination. Keep human authority distinct from agent execution permissions. Share findings with evidence and uncertainty without declaring them policy.]

Approval and changes

Approver: [name and date, or pending]. Anyone may propose a correction within their authorized scope; [owner] approves policy changes.