Motherduck Create Dive logo

Motherduck Create Dive

Organization
motherduckdb
motherduck-create-dive

Create, edit, design, publish, share, or embed MotherDuck Dives, including responsive layout, light/dark themes, and buttons that write back through Flights.

Overview

Publishermotherduckdb
Repositoryagent-skills
Skill namemotherduck-create-dive
Stars
62
Forks
6
Bundled files
4
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 4 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by motherduckdb on GitHub. Read the source before you install it.

Installation

Install the Motherduck Create Dive AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/motherduckdb/agent-skills/tree/main/plugins/motherduck-skills-claude/skills/motherduck-create-dive
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/motherduckdb/agent-skills.git /tmp/agent-skills
mkdir -p .claude/skills
cp -r /tmp/agent-skills/plugins/motherduck-skills-claude/skills/motherduck-create-dive .claude/skills/motherduck-create-dive
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Motherduck Create Dive in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Motherduck Create Dive on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Motherduck Create Dive is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Create and Manage MotherDuck Dives

Source Of Truth

  • Prefer current MotherDuck Dive docs first.
  • Non-negotiable ordering: when MotherDuck MCP is available, call get_dive_guide before writing Dive code and before every save_dive, update_dive, or edit_dive_content. The guide defines the current component API and runtime libraries.
  • get_dive_guide also surfaces relevant conventions from the reserved dives Guide topic. Apply those conventions when they do not conflict with the user's explicit requirements.
  • Use the blessed Dives example repo as the reference implementation for local preview, Dives-as-code layout, metadata, CI previews, and deploy scripts.
  • Use Dives SQL functions when the user wants a scriptable SQL-native create/read/update/delete workflow instead of MCP tools.
  • Treat ordinary Dives and embedded Dives separately. Verify current plan entitlements before promising an embed rollout; do not preserve plan names or availability claims from memory.

Default Posture

  • First classify the job: new Dive, existing Dive edit, Dives-as-code workflow, team sharing, or embedding.
  • Validate the underlying SQL and schema first with motherduck-explore and motherduck-query; a good Dive starts with a correct query.
  • Keep Dive queries fully qualified and SQL-heavy; let React handle presentation, not data reshaping.
  • Treat the component contract as React + useSQLQuery, a default export, supported runtime libraries, explicit loading/empty/error states, and no browser-side secrets.
  • New Dives start as Draft. Promote a validated Dive to Ready only when the requested delivery includes publication. Only an admin can mark a Dive Endorsed; never self-endorse an agent-created Dive.
  • When reusing existing work, prefer Endorsed and then Ready Dives. Archived Dives are retired and excluded from default agent listings unless explicitly requested.
  • When local preview uses REQUIRED_DATABASES, keep the export on one line and mirror the real share dependencies in metadata or save/update inputs. Avoid aliases that collide with existing database names.
  • Preserve an existing Dive’s visual system for scoped edits. For a new Dive, choose a concrete theme direction that fits its audience.
  • Prefer one query per visual section or interaction surface rather than one giant cross-purpose query.
  • Preview locally before saving when the environment supports it.
  • For existing Dives, read the current content and version metadata before overwriting anything. MCP list_dives returns current_version, and read_dive can fetch historical versions. Use edit_dive_content (exact text replacements) for focused edits and update_dive for a full rewrite. Use view_dive to show the user a saved Dive, and delete_dive only after explicit confirmation; deletion is permanent.
  • Add export controls only when the use case calls for them: exportAs on a query result or the useExport hook (CSV, JSON, Parquet, XLSX), started from a click. Do not build Blob URLs, hidden anchors, or COPY TO exports.
  • Dive SQL is read-only. For actions such as approving a record, submitting a form, or refreshing a pipeline, have a button start a Flight the Dive owner controls with useRunFlight and track it with useFlightRun (from @motherduck/react-sql-query). Hardcode the Flight ID, pass only string config values, disable the control while a run is pending, render the hook's error, and rerun the affected queries when the run succeeds. Viewers need a run grant: recommend the grantees and grant only after the user confirms. Embedded Dives cannot run Flights.
  • Viewers can open AI Chat in the Dive header. It sees the Dive's queries, data, and current filters, and can edit the Dive with each change kept in version history. Clear section titles, query names, and the Dive description make its answers better.
  • Treat embedded Dives as the first-choice path when a product needs a live read-only Dive surface with a backend-created embed session. Move to motherduck-build-cfa-app when the app needs custom API contracts, writes, non-Dive routing, tenant policy enforcement, or richer authorization.
  • For shared repos or CI/CD, use a service-account token so Dive ownership is not tied to one human user.

Feature-Complete Defaults

get_dive_guide owns the runtime contract and the MotherDuck visual rules (palette, chart sizing, no card chrome, static Tailwind only); do not restate them. A new Dive or full redesign should also ship with these on the first save. Start from artifacts/feature_complete_dive.tsx, which implements all of them, and delete what the Dive does not need. For a scoped edit, preserve the existing design and check only the affected states.

  • Reflows from about 320 px to wide panes: breakpoint grids (2x2 KPIs on narrow, one row on wide), ResponsiveContainer width="100%", no fixed widths or page-level horizontal scroll.
  • Light and dark themes from one token object per theme, applied with inline style, starting from the viewer's system preference.
  • Filters, sort, and selected view in useDiveState so they survive refresh and travel with a shared link.
  • Every section has its own loading skeleton, empty message, and role="alert" error.
  • A "data as of" freshness line and plain-language titles with units, so viewers and AI Chat can read the Dive.
  • Export with exportAs when people will take the data elsewhere.
  • When the Dive should change data, a write-back control backed by a Flight instrumented for the Flight Explainer.
  • Styled controls (the runtime strips button styles) with 44 px targets, visible focus, and non-color status cues.
  • Do not call a Dive mobile-friendly from responsive CSS alone; report the viewports and states you checked.

Workflow

  1. Choose the requested delivery path: workspace Dive, existing edit, Dives-as-code, write-back actions, sharing, or embedding. Follow only that path; a chart request does not itself authorize sharing, embedding, or write-back.
  2. Explore the live schema and validate the core SQL first.
  3. Call get_dive_guide if MCP is available, then design the story, sections, interactions, and theme.
  4. Build or edit the Dive component, using local preview/hot reload when possible.
  5. Call save_dive, update_dive, or deploy only after queries, loading states, required resources, and visual behavior are correct.
  6. Read the resulting URL, version, and status back. If the requested delivery includes publication, promote a validated Draft to Ready through the write-capable path.
  7. If teammates or application users need access, check three things: they can view the Dive, every REQUIRED_DATABASES entry is a share they can read, and they hold a run grant on each Flight the Dive runs. Prefer an existing share of an owned database (MD_INFORMATION_SCHEMA.OWNED_SHARES). When the user asks to share the Dive's data with the whole organization, share_dive_data creates org-scoped shares and updates the Dive in one call; use motherduck-share-data for restricted audiences, and the embed-session flow for application users.

For answer, review, or planning requests, stop at the requested design or code artifact. For create, update, or deploy requests, carry the requested in-scope operation through preview and validation; ask before destructive replacement or a broader external rollout that the request did not authorize. Before save_dive or update_dive, follow the review step that get_dive_guide sets for your client: when it asks you to show the final source or preview and get the user's go-ahead, do that once, unless the user already said to save without review.

References

Read only the reference sections needed for the current task.

  • Copy and adapt artifacts/feature_complete_dive.tsx as the starting component for a new Dive.
  • Read references/RESPONSIVE_DIVE_DESIGN_SYSTEM.md for the layout grid, component anatomy, theme tokens, filter model, and design anti-patterns of a new Dive or redesign.
  • Read references/VISUAL_QA_PLAYBOOK.md for the screenshot and review loop when a full design handoff or evidence is requested.
  • Read references/DIVE_DESIGN_GUIDE.md for authoring workflows, useSQLQuery mechanics, Dives-as-code, editing/version history, sharing, embedding, write-back with Flights, Agent Chat, SQL functions, theming prompts, chart-selection rules, loading/error states, layout patterns, and implementation gotchas

Related Skills

Load related skills only for missing capabilities; reuse established context.

  • motherduck-explore for discovering the real tables, views, and dimensions before visualizing them
  • motherduck-query for validating the SQL each Dive section will run
  • motherduck-build-dashboard when the work is really a multi-section dashboard composition problem
  • motherduck-build-cfa-app when the requirement is a fuller product surface with per-customer isolation or backend policy control
  • motherduck-cli when the agent has a shell and should keep Dive source in local files
  • motherduck-create-flight for the Flight a write-back button runs
  • motherduck-manage-guides for reusable personal or organization Dive conventions

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Motherduck Create Dive AI skill do?

Create, edit, design, publish, share, or embed MotherDuck Dives, including responsive layout, light/dark themes, and buttons that write back through Flights.

Why use Motherduck Create Dive on TypingMind?

Because you install it once and use it with any model. Motherduck Create Dive is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Motherduck Create Dive in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/motherduckdb/agent-skills/tree/main/plugins/motherduck-skills-claude/skills/motherduck-create-dive. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Motherduck Create Dive?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Motherduck Create Dive?

As many as you like. As long as a model supports skills, you can use Motherduck Create Dive with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Motherduck Create Dive AI skill free?

Yes. It is published on GitHub by motherduckdb under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇