Case Study: ShynSmart

A vibe-coded marketplace.
Real users.

ShynSmart is a jewelry deal marketplace built with AI tools, live in production, with real users browsing and searching every day. It looked polished. It loaded fast. It worked.

It ran through StackSip anyway. The score came back: High risk; not production ready.Five criticals issues found. None of them visible. All of them dangerous.

Duplicate Catalog Listings

Finding 3: Data poisoning via unique constraint gaps

ShynSmart Homepage

ShynSmart Homepage: A polished production marketplace

Audit Score 45

StackSip Score: 45 (High Risk)

API Cost Risk

Finding 1: Hidden API cost spirals with no limits

Environment Variable Mismatch

Finding 2: Silent failures due to env var mismatch

Duplicate Catalog Listings

Finding 3: Data poisoning via unique constraint gaps

ShynSmart Homepage

ShynSmart Homepage: A polished production marketplace

The bill with no ceiling

The platform used OpenAI to process product listings. StackSip flagged it immediately: no input size limit, no cost cap. One bloated scrape job or a small flood of listings, and the API bill spirals with nothing to stop it. The code looked fine. The risk was completely hidden.

Blank screens, No explanation

StackSip caught two different environment variables pointing to the same Django backend. In production, if either is missing or mismatched, pages silently fail. Blank screens. No error. No clue why. The kind of bug that gets reported by a user, not a log.

The catalog that breaks itself

The deduplication logic looked like it was working. StackSip found the gap: No uniqueness constraint. A URL with a trailing slash? Same deal, saved twice. Users see repeated listings. The marketplace looks broken. Nobody knows why.

None of these were obvious. That's exactly the point.

The app worked. It just wasn't ready. StackSip found what wasn't visible before a real user did.

Try it now →