Dreambase Data Agent Skills: Finally, an AI that doesn't hallucinate your data?

Sup nerds. Sick of AI generating shiny dashboards where the numbers make absolutely zero sense? Yeah, me too. You dump a database into a prompt, and the bot creates a beautiful chart that completely misinterprets what your metrics actually mean. Just saw this thing called Dreambase Data Agent Skills drop on Product Hunt, claiming to fix the whole "AI hallucinating your business logic" mess.

The TL;DR: What witchcraft is this?

Basically, Dreambase launched something called "Skills". No, it's not a dusty config file you update once and forget. It's a composable layer that bundles your data sources (Supabase, Stripe, PostHog, random APIs), your spaghetti business logic, and visualization rules into one neat package.

Instead of explaining your schema in every prompt, you define a "SaaS Subscription Skill" once. From then on, every AI agent or report uses that single source of truth. It stops the AI from blindly guessing that status_id = 3 means "premium user" when it actually means "soft deleted". Oh, and it auto-scans your Supabase schema to prepopulate this stuff during onboarding. They’re also bringing CLI and MCP (Model Context Protocol) access soon.

What’s the peanut gallery saying?

Sitting at 136 upvotes on PH, it didn't break the internet, but it got some solid dev chatter.

• The Founders flexing: Andy and Eduardo are out here holding their ground. They claim this took over a year to build. The coolest part? Skills aren't static; agents read them as context and live-check data schemas to catch drift. If your schema changes, the agent actually realizes it.
• The Pragmatic Skeptics: One dev hit them with the real question: "No ETL sounds dope, but what if my database is a dumpster fire with inconsistent names? Can your auto-scan survive that?" The founder's reply was smooth: They don't just rely on names. They use "Human-in-the-loop" onboarding and AI planning. Translation: You still gotta review its homework initially, but it compounds into a smart semantic layer over time.
• The Hype Boys: Someone just commented "LFG". Classic PH seeder or a really supportive buddy.

The C4F Verdict: Magic or just another wrapper?

Generating charts with AI is cheap. Ensuring the AI actually understands your chaotic, undocumented data model is the real million-dollar problem. Dreambase is tackling this head-on by standardizing the context layer.

Takeaway for the trenches: Stop writing 5-page prompts every time you want an AI to query your DB. Build a semantic layer. AI is only as smart as the context you feed it. If you're running Supabase, this tool might actually save you some grey hairs. Worth a spin, or at least worth stealing the architectural idea for your next side project.

Source: Product Hunt - Dreambase AI