How do I install the Connect Datasource skill?

Run npx skillmds@latest add mindsdb/connect-datasource in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Connect Datasource safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Connect Datasource?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Connect Datasource free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Connect Datasource?

mindsdb (@mindsdb) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Connect Datasource by mindsdb — Agent Skill

Name: Connect Datasource
Author: mindsdb

Connect Datasource

MANDATORY reading before connecting a data source, database, or service through the guided (interactive) flow — when the user wants to add/connect Postgres, MySQL, Snowflake, BigQuery, Gmail, HubSpot, Salesforce, Shopify, or any other service and has NOT pasted credentials directly into chat. Loads the connection tools (lookup_connector, request_credentials, label_connection) and the full Data Vault workflow: registry lookup, rendering the credential form, and labelling saved connections. Recall it when the user asks to connect/add a data source. (If the user pastes credentials directly, use connect_new_datasource instead — it is always available.)

mindsdb 2a7b887 2.8 KB Updated Sep 17, 2026

File contents

DATA VAULT WORKFLOW — when the user asks to connect a service or database and has not pasted credentials directly:

LOOKUP FIRST. Call lookup_connector with the user's wording (e.g. "google mail", "my postgres database"). The returned form blob is the SAME spec the in-app Connector Picker uses — the registry already encodes OAuth flows, methods[], how_to markdown, help URLs, and the method-picker UI for services like Gmail. Handcrafting a form from memory produces a strictly worse result, so always look up first.
RENDER THE FORM. Pass the looked-up form spec to request_credentials VERBATIM — tweak only selected_method or subtitle, and copy _connector_id onto the spec. Never strip or paraphrase methods[], OAuth blocks, how_to, or help_url. Include the markdown block the tool returns VERBATIM in your next message (blank lines around the fence) so the form renders in the side panel.
- Handcraft a spec ONLY when the registry returns no match, using your own knowledge of the service's auth shape (host/port/user/password, API key, or OAuth). For engines with several auth options emit methods[] instead of fields[], mark the simplest recommended: true, and pre-set selected_method if the user already signalled a preference.
STOP THERE. The server tests the connection and saves credentials on submit. Your job is done — do NOT call request_credentials again unless the user asks to connect a different service or explicitly requests a new form.
LABEL WHEN ASKED. Once a connection is saved and the user clarifies which account is which (e.g. two Gmail accounts, and support@… is the support address), call label_connection(engine=…, name=<slug>, label=…). The label shows beside the connection in Connected Data Sources so the right account can be picked later. Never guess a label — ask the user first, then persist it.

STRICT: field VALUES never appear in chat — don't echo them, don't include them in any form spec, don't paraphrase them. The chat-emitted form must FEEL identical to the in-app Connector Picker; the registry lookup is what guarantees that parity.

---name: connect-datasourcedescription: MANDATORY reading before connecting a data source, database, or service through the guided (interactive) flow — when the user wants to add/connect Postgres, MySQL, Snowflake, BigQuery, Gmail, HubSpot, Salesforce, Shopify, or any other service and has NOT pasted credentials directly into chat. Loads the connection tools (lookup_connector, request_credentials, label_connection) and the full Data Vault workflow: registry lookup, rendering the credential form, and labelling saved connections. Recall it when the user asks to connect/add a data source. (If the user pastes credentials directly, use connect_new_datasource instead — it is always available.)---DATA VAULT WORKFLOW — when the user asks to connect a service or database and has not pasted credentials directly:1. LOOKUP FIRST. Call `lookup_connector` with the user's wording (e.g. "google mail", "my postgres database"). The returned `form` blob is the SAME spec the in-app Connector Picker uses — the registry already encodes OAuth flows, `methods[]`, `how_to` markdown, help URLs, and the method-picker UI for services like Gmail. Handcrafting a form from memory produces a strictly worse result, so always look up first.2. RENDER THE FORM. Pass the looked-up `form` spec to `request_credentials` VERBATIM — tweak only `selected_method` or `subtitle`, and copy `_connector_id` onto the spec. Never strip or paraphrase `methods[]`, OAuth blocks, `how_to`, or `help_url`. Include the markdown block the tool returns VERBATIM in your next message (blank lines around the fence) so the form renders in the side panel. - Handcraft a spec ONLY when the registry returns no match, using your own knowledge of the service's auth shape (host/port/user/password, API key, or OAuth). For engines with several auth options emit `methods[]` instead of `fields[]`, mark the simplest `recommended: true`, and pre-set `selected_method` if the user already signalled a preference.3. STOP THERE. The server tests the connection and saves credentials on submit. Your job is done — do NOT call `request_credentials` again unless the user asks to connect a different service or explicitly requests a new form.4. LABEL WHEN ASKED. Once a connection is saved and the user clarifies which account is which (e.g. two Gmail accounts, and `support@…` is the support address), call `label_connection(engine=…, name=<slug>, label=…)`. The label shows beside the connection in Connected Data Sources so the right account can be picked later. Never guess a label — ask the user first, then persist it.STRICT: field VALUES never appear in chat — don't echo them, don't include them in any form spec, don't paraphrase them. The chat-emitted form must FEEL identical to the in-app Connector Picker; the registry lookup is what guarantees that parity.

Frequently asked questions