What are you trying to do?
Choose your situation. Open a tool, then copy its prompt into your AI conversation.
I’m about to ask AI somethingAsk Better QuestionsOpen toolClose tool
AI is more useful when the task, context, and constraints are clear.
Start by identifying:
- What you are trying to accomplish
- The relevant context
- The constraints
- What a useful result should look like
Copy this prompt
Help me clarify this task before solving it. Restate the goal, identify missing context or constraints, and ask only the most important follow-up questions.
Use it when: The request feels vague, overloaded, or likely to produce a generic answer.
Limitation: A clearer question improves the odds of a useful response, but it does not guarantee that the answer will be correct.
AI already gave me an answerCheck the AnswerOpen toolClose tool
A confident answer is not the same as a verified answer.
Check:
- What is directly supported
- What is inferred
- What is uncertain
- What depends on current or external information
- What should be checked independently
Copy this prompt
Review your answer and separate the claims into: supported facts, reasonable inferences, assumptions, and items that require external verification. Then tell me what evidence would change the conclusion.
Use it when: The answer includes factual claims, recommendations, estimates, or anything that could affect a real decision.
Limitation: AI can misjudge its own accuracy. Important claims still need to be checked against reliable sources or direct evidence.
I have an idea or planWhat Could Go Wrong?Open toolClose tool
A promising idea can still fail because of hidden assumptions, missing constraints, or second-order effects.
Look for:
- What has to be true for this to work
- What could break
- Who might use it differently than expected
- What burden it creates
- What happens if the underlying conditions change
Copy this prompt
Challenge this idea. Identify the strongest assumptions, likely failure modes, hidden costs, affected stakeholders, and conditions that would make the idea unsafe, ineffective, or not worth pursuing.
Use it when: An idea feels exciting, obvious, or ready to act on.
Limitation: Red teaming can expose risk, but it cannot prove that an idea will fail or succeed.
Something isn’t workingWhat’s Actually Causing the Problem?Open toolClose tool
Work backward from the visible problem before fixing the wrong thing.
Check each layer:
- The information or evidence
- The tool or environment
- The instructions or rules
- The workflow or handoffs
- The reasoning or diagnosis
Copy this prompt
Help me diagnose this problem before proposing a fix. Separate the visible symptom from possible causes in the information, tools, instructions, workflow, handoffs, permissions, environment, and reasoning. Rank the most plausible causes by evidence and tell me the cheapest useful test for each.
Use it when: The obvious explanation may only describe where the problem appeared.
Limitation: A plausible diagnosis is still a hypothesis until evidence distinguishes it from alternatives.
More toolsHide extra tools
Did It Actually Work?Open toolClose tool
Check whether a finished-looking task produced the result you wanted.
Copy this prompt
I’m going to show you a task and the evidence that it is finished. Identify where the intended result actually needs to exist. Separate proof that work was performed from proof that the receiving person or system got the intended outcome, then tell me what still needs to be verified.
Use it when: A task is marked done, but you need evidence that the intended person or system can use the result.
Limitation: A completed file or successful test proves only what it checked; verify the outcome where it matters.
Are You Measuring the Right Thing?Open toolClose tool
Check whether a target or score can improve while the real result gets worse.
Copy this prompt
I’m going to give you a goal and the way I’m measuring success. Look for ways the measurement could improve while the real result gets worse. Identify what behavior the current target rewards, what important value it may miss, and what I should measure or check alongside it.
Use it when: You are setting a target, score, quota, or measure of success.
Limitation: A suggested replacement measure can also mislead. Check real outcomes alongside the numbers.
Where Is the Handoff Breaking?Open toolClose tool
Find what gets lost when work moves between people, tools, or steps.
Copy this prompt
Map this process as a series of handoffs. At each transition, identify what information, responsibility, authority, state, evidence, or next action has to survive. Show me where something can be dropped, misunderstood, duplicated, delayed, or left without an owner.
Use it when: Work breaks as it moves between people, tools, conversations, or steps.
Limitation: A process map is a hypothesis. Check the actual handoff with the people or systems involved.
What Might Be Out of Date?Open toolClose tool
Find assumptions and information that may no longer match the current situation.
Copy this prompt
Review this plan, answer, or decision and identify the parts that depend on information or assumptions that may have changed. Separate what still appears stable from what needs a current check, what may no longer apply, and what is unknown. Prioritize checks that could actually change the decision.
Use it when: You are reusing a plan or answer that depends on changing information or conditions.
Limitation: AI may not have current sources. Treat items as needing verification until you check; old does not automatically mean wrong.
What’s Still Not Done?Open toolClose tool
Check a project or plan for work that disappeared, got skipped, or only looks finished.
Copy this prompt
Compare this status update with the original goal and scope. List what is done, what is still executable, what is blocked, what is intentionally deferred, and anything that disappeared from the plan without an explicit decision. Do not treat a completed subtask as proof that the larger goal is complete.
Use it when: You are reviewing a status update, closing a project, or resuming unfinished work.
Limitation: Missing original requirements limit the review. Do not treat an unmentioned task as completed or authorize new work by inference.