Skip to main content
skill CX Codex

live-artifact — Claude Code skill: what it does

live-artifact builds refreshable HTML artifacts from project or connector data. Use it for live dashboards, reports, and synced data views.

Apache-2.0 123 installs ★ 108,855 updated 2026-10-25

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add nexu-io/open-design --skill live-artifact -a claude-code
Open on GitHub

What live-artifact does

This skill builds an OpenDesign artifact whose presentation can be refreshed as its data changes. It writes template.html, data.json, artifact.json, and provenance.json; the daemon then derives index.html. Template bindings handle scalar values and one-level list repetition with data-od-repeat.

The workflow accepts local/project sources and daemon connector tools. It keeps preview data compact, records source and transformation notes, and registers or updates the artifact through OpenDesign wrapper commands. Stored JSON must exclude credentials, API keys, cookies, authorization headers, and raw provider payloads.

The skill targets Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and OpenCode. It requires shell and file_write capabilities. index.html is generated output, so the source should be template.html together with data.json.

What it helps with

  • build a refreshable HTML dashboard from project data
  • turn source data into a reusable report template
  • connect a view to local data or a daemon connector
  • record source provenance and transformation notes

Example prompts for the agent

  • Create a live dashboard from the project data and make its template.html refreshable.
  • Build a report that can later pull data through a connected daemon connector.
  • Create a synced data view and record its source details in provenance.json.

How to install live-artifact

1Skills CLI (Vercel)

npx skills add nexu-io/open-design --skill live-artifact -a claude-code

2Manually: put the files in place

~/.claude/skills/live-artifact/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/live-artifact/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about live-artifact

How do I install live-artifact in Claude Code?

Run npx skills add nexu-io/open-design --skill live-artifact -a claude-code. The skill file will be at ~/.claude/skills/live-artifact/SKILL.md.

What does live-artifact do?

It creates an OpenDesign HTML artifact with a data-backed template, preview data, and source metadata. The presentation can be refreshed later without redesigning it.

Which files does live-artifact use?

The source files are template.html, data.json, artifact.json, and provenance.json. The daemon derives index.html from the template and data.

Source text

SKILL.md · Apache-2.0
---
name: live-artifact
description: |
  Create refreshable, auditable OpenDesign artifacts backed by connector or local data.
  Trigger when the user asks for live dashboards, refreshable reports, synced views, or reusable data-backed artifacts.
triggers:
  - "live artifact"
  - "live dashboard"
  - "refreshable dashboard"
  - "live report"
  - "refreshable report"
  - "synced view"
  - "可刷新"
  - "实时看板"
od:
  mode: prototype
  scenario: live
  preview:
    type: html
    entry: index.html
    reload: debounce-100
  design_system:
    requires: true
  outputs:
    primary: index.html
    secondary:
      - template.html
      - artifact.json
      - data.json
      - provenance.json
  capabilities_required:
    - shell
    - file_write
---

# Live Artifact Skill

Create an OpenDesign live artifact: a project-scoped, previewable HTML artifact whose data can later be refreshed without redesigning the presentation.

## Resource map

```
live-artifact/
├── SKILL.md
└── references/
    ├── artifact-schema.md      ← `references/artifact-schema.md`: artifact files, DTO shape, template binding rules
    ├── connector-policy.md     ← `references/connector-policy.md`: connector safety, redaction, credential boundaries
    └── refresh-contract.md     ← `references/refresh-contract.md`: source metadata, refresh execution, snapshots
```

## Current status

Use the references in this directory as the source of truth for the live artifact file contract. Prefer daemon wrapper commands over raw HTTP when registering or updating live artifacts.

## When to use this skill

Use this skill when the user asks for a data-backed view that should remain useful after the first render, for example a live dashboard, refreshable report, synced status page, auditable data view, or artifact that can later be refreshed from local/project data or connectors.

Before creating files, decide whether the user actually wants a live artifact or a normal static artifact:

- Use a live artifact when the user mentions refresh, sync, recurring updates, connector-backed data, source/provenance tracking, dashboards, reports, or reusable data-backed views.
- Use a normal static artifact when the user only wants a one-off HTML/mockup/image/file and does not need refresh, source metadata, or data/provenance panels.
- If the intent is ambiguous, ask one short question: “Should this be refreshable/live, or just a static artifact?”

## Workflow

1. **Resolve scope and data source without blocking on connected connectors**
   - Identify the preview goal, audience, data freshness expectations, and whether refresh should be possible later.
   - If the user explicitly names a connector/source such as Notion, GitHub, Slack, or Google Drive, do not ask “where should the data come from?” before checking daemon connector tools.
   - Prefer local/project sources or daemon connector tools when available.
   - Do not call provider APIs directly when a daemon connector/wrapper exists.
   - If connector data is needed, first list connectors with `"$OD_NODE_BIN" "$OD_BIN" tools connectors list --format compact`. If the named connector is present with `status: "connected"`, choose an appropriate read-only `auto` tool from its catalog and execute it through the connector wrapper.
   - For Notion specifically, a connected `notion` connector plus a user brief that names Notion is enough to start with `notion.notion_search` using a query derived from the requested artifact/topic. Use `notion.notion_fetch_database` only when the user supplied a database id or the search result clearly identifies one.
   - Ask the user a data-source question only when no matching connected connector exists, multiple connected candidates fit equally well, or the requested artifact has no usable topic/query to search for. If you must ask, be specific: ask for the page/database/topic or permission to search broadly, not “where is the Notion data source?”

2. **Author the source files**
   - Write `template.html` as the human-designed HTML template. The daemon hydrates it with `data.json` using the `html_template_v1` binding contract — stay inside it or hydration fails and the raw template ships with visible `{{…}}` tokens:
     - Scalars: `{{data.path.to.value}}`. Paths start with `data`, dot-separated; numeric array indexes are allowed (`{{data.kpis.0.value}}`). Every binding must resolve to a single string/number, not an object or array.
     - Lists: repeat one element with `data-od-repeat="item in data.items"`, then bind loop-scoped fields inside it as `{{item.label}}` (or `{{item}}` for a scalar array). One level only; `data.*` still works inside the repeat for globals.
     - Do NOT hand-loop in a `<script>` (scripts are stripped from previews), and do NOT reference a bare loop variable like `{{metric.value}}` without a matching `data-od-repeat` scope — that binding is unresolvable. Nested repeats, conditionals, filters, helpers, and raw/triple-brace interpolation are unsupported.
   - Write `data.json` as the canonical preview data used by those bindings — real values, not placeholders.
   - Write `artifact.json` with the live artifact metadata, preview declaration, document declaration, and safe source descriptors.
   - Write `provenance.json` with concise source notes, timestamps, non-sensitive connector references, and transformation notes.
   - Do not author `index.html` as source. The daemon derives `index.html` from `template.html` and `data.json`.

3. **Keep data compact and preview-oriented**
   - Store only normalized values needed by the preview.
   - Summarize large lists, provider responses, or logs before writing them into `data.json`.
   - Stay within the bounded JSON rules in `references/artifact-schema.md`.

4. **Apply safety rules before registration**
   - Never store credentials, OAuth tokens, API keys, cookies, auth headers, raw provider responses, HTTP envelopes, full payloads, or secret-like fields in `artifact.json`, `data.json`, `provenance.json`, or source metadata.
   - Avoid forbidden key names such as `raw`, `rawResponse`, `payload`, `body`, `headers`, `cookie`, `authorization`, `token`, `secret`, `credential`, and `password` anywhere in persisted JSON.
   - Use escaped `html_template_v1` interpolation only. Raw/unescaped HTML interpolation is not allowed.

5. **Register or update through daemon wrappers**
   - Use the OpenDesign daemon wrapper commands via `"$OD_NODE_BIN" "$OD_BIN"` instead of raw `curl`, bare `node`, or bare `od`:

     ```bash
     "$OD_NODE_BIN" "$OD_BIN" tools live-artifacts create --input artifact.json
     "$OD_NODE_BIN" "$OD_BIN" tools live-artifacts list --format compact
     "$OD_NODE_BIN" "$OD_BIN" tools live-artifacts update --artifact-id "$ARTIFACT_ID" --input artifact.json
     ```

   - The wrapper reads injected `OD_NODE_BIN`, `OD_BIN`, `OD_DAEMON_URL`, and `OD_TOOL_TOKEN`; do not print, persist, or override token values.
   - Do not include or invent `projectId`; the daemon derives project/run scope from the token.
   - Use raw HTTP only for daemon development/debugging when explicitly requested.

6. **Use connector wrappers for connector data**
   - Discover available connectors and tools:

     ```bash
     "$OD_NODE_BIN" "$OD_BIN" tools connectors list --format compact
     ```

   - Execute a read-only connector tool with a JSON object input file:

     ```bash
     "$OD_NODE_BIN" "$OD_BIN" tools connectors execute --connector "$CONNECTOR_ID" --tool "$TOOL_NAME" --input input.json
     ```

   - Persist only the compact normalized fields needed by the preview plus non-sensitive connector references (`connectorId`, `toolName`, `accountLabel`). Never persist connector credentials, transport metadata, or raw provider output.
   - Do not ask for connector secrets or duplicate setup. If `status` is `connected`, use the listed tools; if it is not connected, tell the user to connect it in the UI.
   - See `references/connector-policy.md` for listing/execution and credential boundaries, and `references/refresh-contract.md` for read-only refresh source metadata.

7. **Report concise results**
   - On success, return the artifact ID/title and note that `index.html` is daemon-derived.
   - On validation failure, fix the source files and retry through the wrapper. Do not bypass validation.

## Required files

Every live artifact creation flow must produce these source files before registration:

- `template.html` — declared skill output and source template for the preview.
- `data.json` — compact, canonical preview data.
- `artifact.json` — create/update input for daemon validation.
- `provenance.json` — safe source and transformation summary.

`index.html` is the primary preview entry declared in frontmatter, but it is derived daemon output rather than agent-authored source.

© nexu-io · Apache-2.0 · nexu-io/open-design

Similar skills

All skills for Frontend and UI →

From the same set140

The whole set →
web-prototype-taste-editorial web-prototype-taste-editorial — a skill for single-page prototypes styled like Notion, Linear or premium docs: warm monochrome, serif display and grotesque body, hairline borders, muted pastel chips, generous whitespace. CX Codex +1 172installs web-prototype-taste-soft web-prototype-taste-soft guides coding agents to generate Apple-like web prototypes. Enforces double-bezel cards, cream canvases, button-in-button CTAs, and spring transitions. CX Codex +1 166installs web-prototype-taste-brutalist web-prototype-taste-brutalist — a skill that teaches the agent to build prototypes in Swiss industrial print style: newsprint canvas, heavy grotesque, hazard red, zero radius, hairline grid. CX Codex +1 165installs html-ppt-taste-editorial html-ppt-taste-editorial builds 16:9 HTML presentation decks in editorial print style: warm cream substrate, three type families, hairline rules, one muted accent. CX Codex +1 162installs html-ppt-taste-brutalist html-ppt-taste-brutalist is a Claude Code skill that shapes HTML presentations in a dark CRT-terminal style. It activates for incident reports, security reviews, and infrastructure walkthroughs. CX Codex +1 159installs open-design-landing-deck open-design-landing-deck turns a JSON brief into a horizontal HTML deck: one screen per slide, arrow keys, swipe, ESC for a thumbnail grid. For pitches, talks and reviews. CX Codex +1 139installs kami-deck The kami-deck skill turns a study into an HTML lab-meeting deck with slide structure, findings, caveats, and a next experiment. It suits research briefings. CX Codex +1 133installs kami-landing kami-landing generates a single-page HTML document styled after kami paper: warm parchment canvas, ink-blue accent, serif at one weight, zero dependencies. CX Codex +1 132installs

Project sponsors

Companies that keep this analytics open