What it looks like on the desk.

The duty editor makes the call in the tool they already use. The hold is written to the story, with the rule it was made against and the person who made it.

Every subscribed tool sees it at once and applies the rules its own newsroom wrote. The graphic does not play. The ticker drops the figure. The clip cannot run. The web desk sees the figure marked rather than missing.

Nothing else. They made the call once, in one place, and did not spend the next hour telling six teams about it.

When the second source lands, the hold lifts in the same place and everything that was waiting is released together, with the clearance attributed and timed.

The change goes to the story once, at source, with what it was and what it now is.

Every tool holding something built on the old fact sees that it is stale. The scripts, the graphic, the social cutdown and the page each raise it in their own way, and each decides for itself whether to hold, revise or withdraw.

Reviews a list of what is affected rather than trying to remember it. The judgement stays human; the recall does not.

A trail showing what was live on the old fact, for how long, and what each tool did about it.

The clip becomes an asset on the story with whatever came with it: where it came from, what credentials it carries, what is missing.

Verification tools read the story and post what they find as assertions: an authenticity check, a match against known material, a transcript. None of them publish anything and none of them act.

Decides. The verdict from a tool is evidence for that decision, never the decision itself, and the gate on the clip reads the verdict where it sits.

The reason the clip ran, or did not, on the record, and attached to the clip rather than to a chat thread.

The clearance sits on the link between the asset and the destination it was cleared for, not on the asset itself.

Each destination reads the clearance for its own route. What was cleared for broadcast does not silently become cleared for everywhere.

Clears the new route, usually in seconds, because the evidence is already there. If the words have changed since the original approval, the gate reopens rather than inheriting it.

Reach without laundering: one decision reused deliberately, not a permission that leaked.

Nothing new. The story already carries what was decided, by whom, against which rule, and which frames each decision applied to.

The tools that were involved each published what they did at the time, in their own message family, on the same story.

Answers the question from the record instead of reconstructing it from six systems and the memory of whoever was on shift.

This is the part that is hard to retrofit and almost free once the context travels: the audit trail is a by-product of how writes work, not a system somebody had to build.

It subscribes like any other participant. It reads the story, and the Skills your newsroom has written are the rules it is held to.

It declares what it finds, in its own family, attributed to it. It never sends work to another tool and never commands anything. If it has nothing useful to say, silence is a valid outcome.

Approves, edits or ignores. Anything an agent generates that reaches an audience becomes an asset with its authorship carried, so it can be told apart later.

An agent that is accountable in the same terms as everything else on the desk, rather than a clever black box nobody can audit afterwards.