Role & context
Step 1Task & reasoning mode
Step 2Standard CoT
Reversed CoT
Devil's advocate
Branching
Meta CoT
Confidence-gated
Creative modifiers
Step 3Output format
Step 4Generated prompt
Fill in the fields above to generate your prompt...
Master prompt template
ROLE: You are a [expert type] specializing in [domain].
CONTEXT: [Describe the situation, background, and key constraints in 2–3 sentences.]
TASK: [State what you want clearly. One sentence.]
REASONING MODE: [Pick one:
- Think step by step.
- Start from the opposite conclusion and work backwards.
- After each step, argue why it might be wrong before proceeding.]
DIVERGENCE: Generate [N] completely different framings of the answer. Each must contradict the previous framing's core assumption.
CONSTRAINT: [Apply one creative rule, e.g. "Use only analogies" / "Under 80 words" / "No adjectives."]
PERSPECTIVE: Answer three times — as a [role A], a [role B], and a [role C].
AMBIGUITY: [Choose: Resolve cleanly / Hold the tension open / End with an open question for the reader.]
OUTPUT FORMAT: [Prose / List / Table / Dialogue / Narrative / Code]
High-leverage research template
You are a senior research analyst with expertise in [field].
I need to understand [topic] for [purpose].
First, assume the mainstream consensus on this topic is WRONG. What evidence supports that? (2–3 points)
Then steelman the mainstream view. What's the strongest case for it?
Finally, synthesize a third position that neither side has articulated.
Constraint: No hedging language. Every sentence must take a clear position.
Format: Three labeled sections — Contrarian Case / Mainstream Case / Synthesis.
Multi-perspective decision template
Context: [Describe the decision you face and what's at stake.]
Analyze this decision from three distinct lenses:
1. PESSIMIST: What's the most plausible way this goes badly? Be specific.
2. OPTIMIST: What's the most plausible path to this working well?
3. CONTRARIAN: What would someone argue who thinks both of the above are asking the wrong question entirely?
After all three, identify the single biggest hidden assumption shared by all three perspectives.
Do not resolve which view is correct. Hold the tension.
Creative divergence template
Task: [State the creative or analytical challenge.]
Generate 5 completely different approaches to this. Rules:
- Each approach must start from a different first principle
- Approach 2 must directly contradict Approach 1's core assumption
- Approach 5 must be the one you think I would least expect
Label each: [Approach name] + [Core assumption it starts from] + [The approach itself]
Constraint: Use only analogies to explain each approach — no direct descriptions.
Self-auditing meta-prompt
You just produced an output in response to a prompt. Now audit it:
1. SPECIFICITY: What is the most ambiguous word or phrase in your output? Define it precisely.
2. CONSTRAINT CHECK: What constraint in the original prompt, if removed, would have changed your output most?
3. DIVERGENCE: If you had to rewrite this output from a completely opposite starting assumption, what would change?
4. SURPRISE: Rate your own output 1–5 on unexpectedness. Justify the rating.
5. HIDDEN ASSUMPTION: Identify the single largest unexamined assumption embedded in your reasoning. State it explicitly.
6. TENSION: What did you resolve that should have been left open?
Do not defend your original output. Criticize it as if you were a skeptical editor.
Snippet library
Click to copyTwisted chain-of-thought patterns
Click to copyPerspective lenses
Verification checklist
1
Specificity test
Could 10 different people interpret this prompt identically? If not — tighten role, context, or output format until they could.
2
Constraint pressure test
Remove one constraint. Does the output become generic? If yes — that constraint is load-bearing. Keep it. Consider doubling it.
3
Divergence audit
Run the prompt 3 times. Are outputs substantially similar? If yes — add "each approach must contradict the last" to force genuine divergence.
4
Surprise quotient
Rate each output 1–5 on unexpectedness. Below 3? The prompt is confirming your priors. Inject a contrarian forcing function or reversed CoT.
5
Ambiguity preservation check
Did the model resolve everything when you wanted open-endedness? Explicitly add: "Do not resolve — hold the tension."
6
CoT integrity scan
Find the step in the reasoning where a hidden assumption was made. Feed that assumption back as a new prompt challenge.
Self-auditing meta-prompt
You just produced an output in response to a prompt. Now audit it:
1. SPECIFICITY: What is the most ambiguous word or phrase in your output? Define it precisely.
2. CONSTRAINT CHECK: What constraint, if removed, would have changed your output most?
3. DIVERGENCE: If you rewrote this from a completely opposite starting assumption, what would change?
4. SURPRISE: Rate your own output 1–5 on unexpectedness. Justify the rating.
5. HIDDEN ASSUMPTION: Identify the single largest unexamined assumption in your reasoning.
6. TENSION: What did you resolve that should have been left open?
Do not defend your original output. Criticize it as a skeptical editor.