A practical method for your work
Prompt Debugging Checklist
Diagnose why an AI answer was generic, wrong, unstructured, or hard to reuse.
Core idea
Bad AI output usually fails for a specific reason: missing goal, thin context, vague output shape, no examples, weak constraints, or no verification step. This checklist turns “the model is bad” into a repair loop you can reuse.
Why it works
LLMs respond to the task interface you give them. Debugging the prompt by component makes failure visible: intent, context, output contract, examples, constraints, evidence, and review criteria can each be fixed separately.
This answer is bad. Make it better.
Debug this prompt before rewriting. Check: 1) goal clarity, 2) missing context, 3) output contract, 4) examples/rubric, 5) source or verification needs, 6) constraints. Return the top 3 causes, then a repaired prompt.
Customize it
Working template
Goal: [what I am trying to accomplish]
Context: [background, audience, constraints]
Use this pattern: Prompt Debugging Checklist
Variables: Original prompt, Bad output sample, Desired output, Missing context, Output contract, Verification rule, Reuse note
Return: [exact output format]
Before finalizing: state limits and one improvementOperating recipe
- Start with the weak version so you know what problem you are fixing.
- Add the missing variables instead of making the instruction longer randomly.
- Ask the model to follow the output contract exactly once.
- Review the first answer against the checklist below.
- Save the improved version as your reusable pattern.
Copy-ready 7-step prompt debug
Use this whenever an answer feels generic, wrong, too long, or hard to reuse. Paste the checklist under your weak prompt, then ask the model to return the top causes and a repaired version.
Check: What result should this prompt produce, and what decision or action will it support?
Repair: Rewrite the first line as: “Help me accomplish [goal] for [audience/context].”
Check: What does the model need to know that is not already in the prompt?
Repair: Add audience, constraints, prior attempts, data/source boundaries, and examples of what “good” looks like.
Check: Did I specify the exact format, sections, length, language, and tone?
Repair: Give a visible contract: headings, bullet count, table/no-table rule, JSON shape, or final checklist.
Check: Did I show a good sample or define how the answer will be judged?
Repair: Add one positive example, one anti-example, or a 3-point rubric before asking for the final answer.
Check: Does this task require current facts, sources, calculations, or tool-based verification?
Repair: Ask the model to label assumptions and separate verified facts from guesses; use tools for anything factual or current.
Check: What must the answer avoid, preserve, or not change?
Repair: Name the hard constraints: budget, file scope, forbidden claims, style rules, safety boundaries, or no-go options.
Check: Can this repaired prompt become a reusable pattern next time?
Repair: Replace one-off details with variables like [audience], [source], [output format], and [quality bar].
1. Goal check
- Check: What result should this prompt produce, and what decision or action will it support?
- Repair: Rewrite the first line as: “Help me accomplish [goal] for [audience/context].”
2. Context check
- Check: What does the model need to know that is not already in the prompt?
- Repair: Add audience, constraints, prior attempts, data/source boundaries, and examples of what “good” looks like.
3. Output contract check
- Check: Did I specify the exact format, sections, length, language, and tone?
- Repair: Give a visible contract: headings, bullet count, table/no-table rule, JSON shape, or final checklist.
4. Example / rubric check
- Check: Did I show a good sample or define how the answer will be judged?
- Repair: Add one positive example, one anti-example, or a 3-point rubric before asking for the final answer.
5. Evidence check
- Check: Does this task require current facts, sources, calculations, or tool-based verification?
- Repair: Ask the model to label assumptions and separate verified facts from guesses; use tools for anything factual or current.
6. Constraint check
- Check: What must the answer avoid, preserve, or not change?
- Repair: Name the hard constraints: budget, file scope, forbidden claims, style rules, safety boundaries, or no-go options.
7. Reuse check
- Check: Can this repaired prompt become a reusable pattern next time?
- Repair: Replace one-off details with variables like [audience], [source], [output format], and [quality bar].Quality checklist
- Did I give the model the real goal, not just the task?
- Did I define the output shape before asking for the answer?
- Did I include examples, constraints, or a quality bar?
- Did I ask for limits, uncertainty, or failure cases?
- Can I reuse this as a pattern next time?
Model notes
Strong for long context, critique, and structured writing. Give it clear sections and examples.
Strong for fast iteration and everyday templates. Be explicit about output format and assumptions.
Useful for broad synthesis and Google-adjacent research. Keep source requirements explicit.
Limits
A repaired prompt cannot fix missing facts, inaccessible sources, weak product strategy, or a task that needs real tools. Use this checklist to improve the instruction, then verify important claims separately.
Read next
Turn this into a reusable workflow.
Get the Prompt Debugging Checklist and Solo Builder AI Setup Pack as language-specific .md files.