I will build a customer chat ai with prompt injection guards and a test report


About this gig
I build a chat or inbox assistant designed to stay on your topics, resist manipulation, and only send replies that pass the rules you set. It runs as a small serverless service (Cloudflare Worker or Node).
What you get:
- Assistant service with your FAQ and price list loaded as data
- Guards: input cleaning, prompt-injection pre-filter, signed conversation history, output checks, rate limits (by package)
- Unsubscribe and human-handoff rules
- Red-team script with attack cases and results (attack results are included in every package; written report is Advanced only)
- Deployment notes
Not included: CRM integration (quoted separately), 24/7 monitoring, legal review of your privacy policy.
Disclosure: conversations are processed by a third-party model API. You decide which provider, and I help you word the disclosure for your site.
Message me first with your FAQ and where the assistant will live, so I can confirm the package fits.
Get to know Bao L. Zhou
Supplier verification, due diligence and 1688 seller contact, based in China
- FromChina
- Member sinceJul 2026
- Avg. response time15 hours
Languages
Chinese, English
FAQ
Will it ever invent a price?
Prices are checked against your list before a reply is sent. If it can't match, it hands over to a person.
Which model provider do you use?
Your choice. I give a recommendation after testing on your FAQ.
Do you guarantee it can't be manipulated?
No. You get the attack cases I ran and the results, so you can see what was tested.

