Step-Budget Reasoning Protocol
Solves a hard problem in a limited number of steps with checkpoints and honest self-assessment.
The explanation is in the selected language; the prompt text stays in English.
Why use it
For calculation, logic or planning problems where a model answers fast but unreliably: it forces a step limit and self-checking.
How to use it
State the problem in [PROBLEM], the number of steps in [E.G. 10 STEPS] and the rigor in [QUICK / CAREFUL / PROOF].
Short example
Problem: 3 workers paint a building in 5 hours, how long for 5 workers? Steps: 6. Rigor: CAREFUL.
What to expect
A numbered solution showing remaining budget, a 0 to 1 confidence score after every three steps, then Answer, Assumption and Check.
Precautions and tips
- Redo the calculations yourself and compare with the answer.
- The confidence score is a guess, not a guarantee.
- Do not rely on it alone for serious financial or medical decisions.
# Step-Budget Reasoning Protocol
## Role
You are a careful problem solver who shows structured reasoning and checks its own work.
## Context
- The problem: [PROBLEM]
- Step budget: [E.G. 10 STEPS]
- Required level of rigor: [QUICK / CAREFUL / PROOF]
## Task
Solve the problem in numbered steps within the budget, stating the remaining budget after each step. After every three steps, add a brief reflection on whether the approach is working and score your confidence from 0 to 1. If confidence is low, backtrack and try another approach. Finish with the answer, the main assumption it rests on and a check of the result.
## Output format
Numbered steps with remaining budget, periodic reflections with confidence, then Answer, Assumption, Check.
## Constraints
- Do not hide uncertainty
- Show calculations explicitly
- If the budget is not enough, say what is left unresolvedThe prompt text is the original English and is not translated: paste it into your AI tool as is. Replace [text in square brackets] or CAPITALIZED placeholders with your own details. Always check the answer.
Interactive mode: search, progress and Python exercises.