Dreambase Data Agent Skills - Analytical skills for data agents running on Supabase

We've been working towards Data Agent Skills since we started Dreambase, we knew early on that context and documentation was the key to unlocking accuracy and repeatability, the industry just took a while to coalesce on the right standards and practices.

The hardest problem in AI analytics is not generating a dashboard, it is making sure the AI understands your business well enough to generate the right dashboard, or the right insight, every time, without you explaining your data model on every query.

Skills solve that. Define your business logic once. Every agent, every dashboard, every report uses it consistently from that point forward.

We have been building toward this for over a year. Watching it ship today is one of those moments where the product finally matches the vision we started with. We've got so much more on the way, exposing our agents into your workflows and tools soon!

Hey, Eduardo here, founding engineer at Dreambase.

A few things I'm proud of in this release:

Skills aren't a config file. They're a composable unit that bundles connectors (Supabase tables, Stripe, PostHog, any API or MCP), business logic, formatting rules, and viz preferences. Our agents read Skills as context, loads them manually if selected or on demand depending on user intent and also analyze your data schemas live when necessary to see if there are any misalignments or additional considerations, which is why queries actually match how your business works.

Onboarding scans your Supabase schema and auto-generates a starter set of Skills. From there, the Skill Builder is a plain-language wizard that picks the right tables and connectors for whatever you're trying to track.

Soon, Skills will be accessible via CLI and MCP, so other agents and platforms can consume them too. If you're building agentic tooling on top of Supabase data, this is the part I'm most excited about.

Happy to nerd out on the architecture, the agent stack, or how we handle schema drift. Ask me anything.

How accurate is the onboarding scan auto-population feature for a database with unclear or inconsistent table names that don't follow standard SaaS conventions?

Analytics without ETL or extra setup sounds great… curious how well it holds up with more complex data.

LFG

The "define once, reuse everywhere" approach to business logic makes a lot of sense - tired of re-explaining what my data means every time I ask an agent to build a report. Using Supabase for auth and data in my own SaaS project, and the gap between raw tables and useful dashboards is real. Does the auto-scan onboarding handle Supabase RLS policies, or does it only look at the schema?

Congrats @alwaysunday and team - pumped for you!

skills as a composable layer is smart — way better than the agent reverse-engineering your schema every run. how do skills handle changes when someone alters a Supabase table? do they break loud, or silently drift?