A boundary brief is a short operating document that tells an agent what it may do, what needs approval, what must be avoided, and when to stop.
That order is backwards for personal automation. The more an agent can do across text, browsers, publishing, and human handoff, the more it needs compact operating limits that travel with the task.
It should be short enough to inject into the agent's task state and concrete enough to affect routing: allowed actions, approval gates, blocked actions, sensitive data rules, and handoff triggers.
What the agent can do without asking again.
What must pause for user consent or human review.
When the agent should report instead of improvising.
Define the task, channel, account, and time window. A boundary without scope becomes a vague personality trait.
Name the actions the agent can take autonomously, such as drafting, summarizing, searching, preparing, or saving context.
List actions that require permission: sending external messages, spending money, publishing, changing accounts, or handling sensitive data.
Tell the agent when to halt, ask, hand off, or report uncertainty instead of continuing.
Capture what boundary was used, what decision it changed, and whether the user corrected the action.
They are most useful when agents move between text, browser work, publishing, and handoff without losing the user's limits.
Use text message AI assistant flows to state what the agent may send, draft, or hold.
Pair with computer-use cache so browsing can continue while risky actions stay gated.
In agent website workflows, separate preview-safe work from public release.
Give humans the boundary that shaped the agent's choices.
Is a boundary brief the same as a system prompt?
No. It can feed a system prompt, but it is a user-facing operating brief with scope, approvals, and stop rules.
Should every task have one?
High-impact and cross-channel tasks should. Low-risk drafts can use a default brief.
Where does Supers fit?
Supers is relevant for personal AI agents spanning messaging, browser automation, publishing, and human escalation.
Boundary briefs make user limits portable. They let personal AI agents carry permissions, stop rules, and approval gates into every channel where work happens.
Start with one workflow and one brief. Then reuse the pattern across text, browser, publishing, and handoff paths.