Query cross-connector business data for agents with Dinobase
Use Dinobase to give agents a SQL layer over SaaS APIs, databases, files, and MCP servers so they can answer cross-system business questions with previewed writes.
npx skills add agentskillexchange/skills --skill query-cross-connector-business-data-for-agents-with-dinobase
Approve this skill for operators who need agents to answer questions that span multiple business systems without stitching together many brittle connector calls. The operator installs Dinobase, adds approved sources such as Stripe, HubSpot, Linear, Postgres, parquet files, or read-only MCP tools, syncs metadata, and lets an agent inspect schemas, describe tables, and run SQL queries across connectors. Use this instead of normal product dashboards when the question requires joins across systems, semantic table context, or repeatable agent-side data investigation; for mutations, keep Dinobase’s preview/confirm flow and human approval in place. The scope boundary is governed cross-connector data retrieval and reviewed SQL mutation for agents, not a generic data platform listing or an autonomous business-operations product card.
What this skill actually does
Inputs and prerequisites: Dinobase CLI; approved credentials for target systems; selected agent runtime or MCP client; operator review for any write operation.
Setup notes: Follow the upstream installation path, add connectors with dinobase add or connector create, run dinobase sync, inspect schemas with dinobase describe, and invoke dinobase query from an agent workflow. Use dinobase serve or mcp-config only when the target client uses MCP.
Source and verification boundary: use https://github.com/DinobaseHQ/dinobase as the canonical source before running the workflow; keep connector configuration, generated SQL, query results, and mutation confirmations reviewable against that upstream documentation.
Framework fit: publish this as a Multi-Framework workflow because the upstream explicitly supports CLI-based agent use across Claude Code, Cursor, Codex, Aider-style shell agents, plus MCP clients; the skill is the cross-connector SQL workflow, not any one framework integration.