Where the Gate Sits
The plan gate is not the first thing in a run, and putting it first is the most common way to make it useless. Before it comes a survey stage: read the anchors, measure the neighborhood, check what the topic actually looks like today rather than what memory says. Only then does the gate open, because a plan proposed without that survey is a guess dressed as a decision, and the person approving it has nothing to push against.
After the gate, the run is autonomous until it reports. That asymmetry is deliberate. The expensive conversation happens once, at the point where changing course is nearly free, and then the work proceeds without interruption.
What the Gate Has to Decide
A gate that asks "does this look okay?" produces "looks good" and protects nothing. What makes it real is an explicit list of the decisions that are expensive to reverse. In this workshop that list is: the identifier the work will live under, the kind of thing being made, the angle — the reason it is worth making at all — the structural breakdown and its size, the density per unit, the voice, the placement in whatever catalog or index the result has to appear in, the prerequisites, the difficulty, and the open trade-offs the author cannot resolve alone.
Notice what is absent: no resourcing, no schedule, no estimate of how long the author will take. The only hours the gate names are the reader's, because density is part of shape. Those other questions are real and they belong somewhere else. The gate is about shape, and shape is the thing that cannot be repaired afterward without redoing the work.
The Trade-Offs Section Is the Load-Bearing One
The list above is mostly things the author can propose confidently. The last item is different: it is the author saying "here are the calls I should not make alone." A gate without that section degrades into a presentation, and a presentation gets approved.
The useful form is a decision with a recommendation attached, not an open question. "I plan to redact the numbers because they teach nothing; say so if you would rather match the sibling that publishes them" gives the approver something to disagree with in one word. "How should we handle the numbers?" hands the work back.
The Gate Is a Stage, So It Gets Logged
What was approved, by whom, and on what basis becomes a record at the moment it happens — not a sentence in a final report written hours later. Three separate runs in this workshop's history reached the end and discovered the gate had never been recorded, because the session treated it as a conversation rather than as a stage boundary. The work was fine; the evidence that it had been agreed was missing.