These are internal skills purpose-built for analytical use cases within Dreambase: they're written for and used by our data agents, the Data Engineer, the Analyst, and Business Insights, rather than general-purpose tasks.

Dreambase agents write these skills to create a semantic map between your raw Connectors and a working AI-native data pipeline. It takes your connected Postgres databases, MCP servers, and API endpoints, along with your instructions and preferences for how their tables, columns, tools, and endpoints should be mapped and interpreted, and turns all of it into the schema understanding your data agents use to build datasets and dashboards.

This is the heart and brain of your custom data platform in Dreambase.

What a Semantic Skill contains

• Overview — what this Skill covers and the business context behind it.

• Data Sources — the specific Postgres tables, MCP tools, and API endpoints this Skill maps, a curated slice of everything connected to your workspace.

• Skill Docs — AI-generated, AI-readable reference documentation describing the Skill's data model, generated directly by Dreambase and kept current as the underlying data sources change.

• Semantic Graph — a visual map of how the Skill's tables and fields relate to each other.

• Linked Dashboards — every dashboard currently built on this Skill.

Why Semantic Skills matter

Every dashboard, report, and Business Insight your data agents produce draws on a Skill's mapping instead of raw, unmapped schema. That mapping is what lets the Data Engineer agent build correct pipelines, the Analyst agent build accurate dashboards, and Business Insights answer high-stakes questions like revenue with confidence, all from the same shared, reusable source of truth.

Using a Semantic Skill

• Type / in the homepage composer or a dashboard's Analyst Agent to reference a Skill directly.

• Combine a Skill with specific # table mentions when you need to blend its curated mapping with something more granular.

• Keep a Skill's Data Sources scoped tightly. The narrower and more precise the mapping, the more accurate the pipeline, the datasets, and the dashboards it produces.

Using Semantic Skills from the Dreambase MCP

Agents connected through the Official Dreambase MCP can work with Skills too. list_skills lists the workspace's Skills and get_skill returns one in full: its data sources, linked dashboards, and technical context. Reading a Skill first saves an agent from rediscovering your schema.

create_skill generates a new Skill with AI from a plain-language request (it takes 30 to 90 seconds), and update_skill edits a Skill's name, description, prompt, technical context, or status. Both are limited to workspace owners and admins.