Write a system prompt and guardrails for your AI chatbot

Your chatbot answers questions about competitors, promises refunds you don't offer, or changes tone halfway through a conversation. The usual cause is a vague system prompt with nobody checking the edge cases.

I'll write a system prompt for your bot or agent and test it against real questions.

What you get:
- A system prompt built from your business info, tone and rules
- Clear limits: topics to avoid, what to say when it doesn't know, when to hand over to a person
- Rules for tricky cases such as discounts, complaints, legal or medical questions and attempts to make it ignore its instructions
- A tone section with 3 example replies in your voice
- A test set of 40 questions, including awkward ones, with the bot's answers before and after
- A short note on where a code-level check is safer than a prompt rule (price lookups, for example)
- Versions tuned for the model you use: ChatGPT, Claude, Gemini or an open-source model

I build chatbots and AI agents for businesses, including a gateway with rules and logs in front of several LLM providers, so I know where prompts stop working and code has to take over.

Send me your current prompt if you have one, a few real conversations and your no-go topics. Building or hosting the bot itself is not part of this offer.