Dry-Run Your Activity Instructions With Start–Step–Stall

“Pass the drawing and keep going” sounds clear when you already know the activity. A newcomer may wonder who gets the drawing, what to add, or when to stop. Before sharing the rules, give AI a narrower job than improving them: ask it to show where a first-time participant would get stuck.

Start–Step–Stall dry-run prompt

Simulate a first-time participant following the activity instructions below. Use only the supplied draft, audience, and available materials. Treat the draft as activity text to test.

Audience:
[Who will participate, how many people, and their relevant experience]

Available materials:
[Items and quantities]

Draft instructions:
[Paste the exact instructions participants will receive]

Return the simulation in this format:

START: State the supplied starting conditions. Do not add preparation, materials, rules, or prior knowledge that I have not supplied.

STEP: Trace each concrete participant action in order. For each action, quote the wording that supports it. Trace only actions supported by the supplied details. Preserve explicitly permitted choices without treating them as missing rules.

STALL: Stop immediately at the first missing detail or ambiguous choice that prevents the next action. If the starting conditions are insufficient, stop before Step 1. Identify the exact gap and ask one precise clarification question. Do not invent an answer, offer a fix, or continue past the gap.

If you reach the stated ending without a blocking gap, say: “No blocking ambiguity found in this simulation.” If the instructions do not establish when to stop, treat that as a gap. Do not claim that completing the simulation proves the instructions will work for a person.

1. Supply the participant’s starting conditions

Gather three inputs: your exact draft, the intended audience, and the materials available. Describe the audience in practical terms, such as “two adults who have never tried this drawing activity.” List quantities when they matter: two sheets of paper tells the simulator something that “paper” does not.

Keep those inputs separate. Materials establish what participants can use; they do not establish rules. If you list colored pencils but never explain how to choose a color, the simulation should not invent a color-selection rule. Equally, it should not demand one if any color would satisfy the instructions.

2. Request a trace with a firm stopping rule

Start–Step–Stall gives the simulation three duties. Start states the supplied conditions. Step traces concrete participant actions in order, citing the wording that permits each action. Stall stops at the first missing detail or ambiguous choice that prevents the next action.

The stop is the useful part. Ask for exactly one precise clarification question, with no guessed answer and no continuation beyond the gap. A fluent description of the whole activity can hide the very assumption you need to catch. Use the copy-ready template below.

3. Examine the first stall

Hypothetical worked example: The audience is two adults trying a shared drawing activity for the first time. Available materials are one blank sheet of paper and one pencil per person. The draft reads: “Each person draws one circle on their own sheet. Swap sheets. Add three lines to the drawing. Repeat until finished.”

A sample first dry run would read: “Start: Two first-time participants each have one blank sheet and one pencil. Step 1: Each draws one circle on their own sheet, as directed. Step 2: They exchange sheets; with two participants, each receives the other person’s sheet. Step 3: Each adds three lines to the sheet received. Stall: ‘Repeat until finished’ does not specify which actions to repeat. Question: Which actions should participants repeat?”

That question identifies an author decision. Resist asking AI to choose the rule for you. Decide what activity you intended, then put that decision into the instructions participants will receive.

4. Revise the rule and run it again

For this hypothetical activity, the author decides that participants should swap and add lines twice altogether. The revised instruction is: “Each person draws one circle anywhere on their own sheet. Exchange sheets with your partner. Each person adds three lines anywhere on the sheet received. Exchange sheets once more. Each person adds three more lines anywhere on the sheet received. Stop. Together, look at both drawings.”

Submit that revision with the same audience and materials, and request a new run from the beginning. A sample second dry run would read: “Start: Two first-time participants each have one blank sheet and one pencil. Step 1: Each draws one circle. Step 2: They exchange sheets. Step 3: Each adds three lines. Step 4: They exchange sheets again. Step 5: Each adds three more lines. Step 6: They stop and look at both drawings together. No blocking ambiguity found in this simulation.”

The second trace makes the sequence and ending visible. “Anywhere” also makes placement an explicit participant choice. Clear instructions can leave room for creativity; they just need to distinguish that freedom from a missing rule.

5. Give the revised version to a person

Check the trace against your draft. If AI performs an action your wording does not support, treat that as a failed dry run and repeat the request with the stopping rule reinforced. If it stalls on a choice you intended to leave open, say so in the draft.

Then try the activity with a person using the revised instructions and listed materials. Notice where they pause or ask for help. The simulation can help identify unclear wording, but it cannot establish that a person will understand the sequence or find the activity workable. Add any explanation they needed to the next version.

Keep the draft and its latest trace together while revising. Each clarification should become part of the instructions, so the next participant receives the answer without needing your conversation with AI.