Define your use case and get a complete system prompt, knowledge base strategy, and conversation design for a custom AI assistant.
Paste into any LLM. Use the output to build a Custom GPT, Claude Project, or any AI assistant.
You are an AI product designer who has built custom AI assistants used by thousands of people daily. Assistant purpose: [WHAT SHOULD THIS AI ASSISTANT DO] Target users: [WHO WILL USE IT] Domain expertise needed: [WHAT KNOWLEDGE DOMAIN] Tone: [PROFESSIONAL/CASUAL/TECHNICAL/FRIENDLY] Platform: [CUSTOM GPT/CLAUDE PROJECT/API-BASED/OTHER] Design a complete AI assistant: **1. System Prompt** - Write the complete system prompt (this is the most important part) - Include: role definition, personality, boundaries, output format - Rules for what it should and should not do - Edge case handling instructions - Conversation style directives **2. Knowledge Base Strategy** - What documents/data to upload - How to structure the knowledge for best retrieval - Document formatting best practices - How to handle information gaps - Update and maintenance schedule **3. Conversation Design** - Opening message / greeting - 5 conversation starters to suggest - Question flow (how it should gather information) - Response format templates (for different query types) - Follow-up question patterns - Handoff protocol (when to escalate to a human) **4. Guardrails & Safety** - Topics to decline (with polite refusal scripts) - Hallucination prevention instructions - Source citation requirements - Confidence expression guidelines - Privacy and data handling rules **5. Testing Scenarios** - 10 test prompts covering common use cases - 5 edge case prompts to stress test - Expected vs unacceptable responses - Evaluation criteria **6. Iteration Plan** - How to collect user feedback - Common failure patterns to monitor - A/B testing approach for prompt variations - Version control for system prompt updates The system prompt should be production-ready. Write it as if you're shipping it today.
This prompt produces reliable output because it leverages audience specification and role assignment and sequential task breakdown. Each element gives the AI model additional signal about what quality looks like for this specific task. Your output will be reliable agent workflows with decision logic, error recovery, and clear completion criteria - the difference between useful AI assistance and a response you immediately delete.
These agentic ai tips will help you get stronger results when using "Custom GPT / AI Assistant Builder" and similar prompts in this category.
"Custom GPT / AI Assistant Builder" is particularly useful in these situations. If any of these scenarios sound familiar, this prompt will save you significant time.
When you use "Custom GPT / AI Assistant Builder" with ChatGPT, Claude, or Gemini, here is what to expect in the AI output.
Adapt "Custom GPT / AI Assistant Builder" to your specific situation by modifying these key areas. The more context you add, the better the results.