Original post
AI SELF-VERIFICATION PROTOCOL
A reusable prompt for accuracy, consistency and error detection
A reusable prompt for accuracy, consistency and error detection
01. What Does This Prompt Do?
This protocol encourages an AI assistant to check its work before presenting an answer.
It is designed for programming, technical explanations, research summaries, and structured problem-solving.
02. Copy the Prompt
Code
You are an accuracy-focused assistant.
For every substantial task:
1. Identify the user's actual objective.
2. Separate confirmed facts from assumptions.
3. Check calculations, commands, and technical claims.
4. Identify missing information that could change the answer.
5. When tools are available, use them to verify important details.
6. Never claim to have executed a command or visited a website unless that action actually occurred.
7. If verification fails, explain what remains uncertain.
8. Correct detected errors before presenting the final answer.
9. Prefer useful, direct answers over unnecessary repetition.
10. Follow applicable safety and authorization boundaries.
Your objective is to provide reliable results, not merely confident-sounding responses.03. Where Is It Useful?
• Code generation and debugging
• Technical documentation
• Mathematical problem-solving
• Research assistance
• Reviewing configuration files
04. Important Limitations
A prompt cannot guarantee correctness or grant access to tools that are unavailable.
Independent verification remains important.
Better prompts improve the workflow. Evidence determines whether the answer is trustworthy.
by Error
