Skip to main content
Agent skill

Live Artifact

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.

Get started

Get the complete skill folder, including its scripts and reference files.

When to use this skill

Use Live Artifact when an AI agent needs a reusable SKILL.md workflow for this job: 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.

When not to use it

Skip Live Artifact when the task is outside the coding category, or when a more specific skill in this directory already covers the same workflow with clearer triggers.

How to install

Where will you use it?

Install scope
Claude Code folder setup
mkdir -p ~/.claude/skills/live-artifact
# Save SKILL.md to ~/.claude/skills/live-artifact/SKILL.md
  1. Claude Code personal install: place the skill and its bundled files at ~/.claude/skills/live-artifact/SKILL.md. For a shared project, commit the folder at .claude/skills/live-artifact/ instead.
  2. Current Claude Code detects SKILL.md changes in watched directories during the session. If you just created a top-level skills directory, run /reload-skills. Check that the skill is listed before testing its trigger.
  3. For Claude chat or Cowork in the Desktop app, manage uploaded skills in Customize > Skills; copying files into a local Claude Code folder is not a Desktop install. Other agents may use different install locations.

Full install guide for Claude, Cursor, and Codex →

Try a starter prompt

After installing the skill, paste this starter prompt into your agent to verify it triggers on a small task.

Starter prompt
Use the Live Artifact skill for this task: 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. Confirm any required inputs or files before making changes.

What supports this recommendation?

83/100 provisional content score. This measures documentation, practical detail and source signals. It is not a measured task success rate or a security certification.

Documentation
92/100
Practical detail
92/100
Source evidence
63/100
Source signals
72/100
Model compatibility evidence
  • claudeuntested

    No model-specific signal or recorded compatibility test was found.

  • chatgptuntested

    No model-specific signal or recorded compatibility test was found.

  • geminiuntested

    No model-specific signal or recorded compatibility test was found.

  • copilotuntested

    No model-specific signal or recorded compatibility test was found.

  • llamauntested

    No model-specific signal or recorded compatibility test was found.

  • perplexityuntested

    No model-specific signal or recorded compatibility test was found.

  • mistraluntested

    No model-specific signal or recorded compatibility test was found.

  • grokuntested

    No model-specific signal or recorded compatibility test was found.

Inferred compatibility is based on content signals. Check the app, its available tools and setup requirements before running a skill.

What this skill does

# 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.

Intended uses

  • 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?”
  • 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 "$ODNODEBIN" "$ODBIN" 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.notionsearch using a query derived from the requested artifact/topic. Use notion.notionfetchdatabase 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?”
  • Author the source files
  • Write template.html as the human-designed HTML template. The daemon hydrates it with data.json using the htmltemplatev1 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.

Implementation notes

  • Review the source instructions and adapt inputs before running the workflow.

Try it, then tell us how it went

A copied prompt does not tell us whether the task worked. Report your result after checking the output. We collect your tool and outcome, never your files or prompt.