Prompt compression: the shorter prompt pack
This pack helps you test whether a longer prompt is automatically more precise. It turns the idea into a small decision you can inspect this week, without inventing customer evidence or hiding behind more output.
Start with the evidence
I need to decide whether a longer prompt is automatically more precise.
Review the material I paste below.
Separate observed behaviour, direct statements, opinions, and assumptions.
Quote the evidence behind every conclusion.
Do not suggest solutions yet.
End with the three most important missing facts.
Useful signals to look for:
- Instructions repeat
- The desired outcome appears late
- Success cannot be judged after generation
Run the smallest honest test
The test is to cut a prompt to outcome, constraints, and success criteria, then compare the result. Decide what passing and failing mean before you see the result. Otherwise almost any outcome can be explained as good news after the fact.
Design a seven-day test to cut a prompt to outcome, constraints, and success criteria, then compare the result.
The risky assumption is: a longer prompt is automatically more precise.
Use real behaviour, not hypothetical intention.
Give me the participant, task, evidence to capture, pass condition,
failure condition, and the decision each result should trigger.
Keep the test small. Do not propose a new product or campaign.
Ask for a real timeline
Write five interview questions about a specific past event.
Cover what happened, what they tried, the workaround, the cost,
and what happened afterward.
Do not ask what feature they want or whether they would use an idea.
Add one short follow-up beneath each question.
Read the result without rescuing it
Review these notes as a skeptical product partner.
Create two columns: what happened and my interpretation.
Flag any interpretation unsupported by an observation.
Give the strongest case for continuing and the strongest case for stopping.
Name the one missing fact most likely to change the decision.
Make the decision
Add detail only when the shorter version fails for a specific reason you can name.
Use this short decision note:
- Decision: What happens now?
- Evidence: What observed behaviour supports it?
- Risk: What could still make it wrong?
- Next check: What will you inspect and when?
- Not doing: What attractive option is waiting?
One prompt to run on everything
Help me decide whether a longer prompt is automatically more precise.
1. Separate behaviour, statements, opinions, and assumptions.
2. Quote the evidence behind each important claim.
3. Name the weakest assumption carrying the most risk.
4. Design a seven-day test to cut a prompt to outcome, constraints, and success criteria, then compare the result.
5. Set pass and failure conditions before the test.
6. Give five past-behaviour interview questions.
7. End with a decision note and one thing we are deliberately not doing.
Do not invent customer facts. Mark missing evidence clearly.
Download the runnable R156 pack
Use the safe dry-run first. The starter validates fictional input and returns a reviewable draft; it does not publish, charge, contact users, or change production systems. The ZIP includes the Node service, tests, Docker setup, n8n workflow, GitHub Actions, and examples.