storykit-assets-skill
v2.0.1
Published
Agent skill for mapping authorized user data to relevant StoryKit widgets and returning opaque hosted resources without exposing widget code.
Maintainers
Readme
storykit-assets-skill
Teach an agent to turn authorized structured data into a relevant, hosted StoryKit widget. The skill directs the agent to search by intent, obey the selected data schema, protect sensitive inputs, and return the MCP resource link without fetching or reproducing widget implementation code.
Install
npx storykit-assets-skill installThis installs SKILL.md at .claude/skills/storykit-data-widgets/SKILL.md. Reload skills and connect
the free StoryKit MCP endpoint:
claude mcp add --transport http storykit https://asset.storykit.space/api/v1/mcpFor another agent or directory:
npx storykit-assets-skill install --dir ./skills/storykit-data-widgetsThe current hosted skill is also available at
asset.storykit.space/skill/SKILL.md.
What the skill enforces
- Use
find_widgetsfor safe discovery and compare the returned data schemas. - Never invent readings or force rows into an unrelated schema.
- Ask for consent before transmitting health, location, account, device, or personal data.
- Minimize rows and exclude credentials, access tokens, and direct identifiers.
- Use
render_widgetand present its expiring resource link; never request or output source/snippets. - Treat the hosted URL as a secret because anyone with it can view the rendered data.
- Create a new immutable render when data changes.
StoryKit's public service requires no API key. It validates bounded payloads, encrypts them at rest, and deletes renders after their configured lifetime.
MIT. Created with love by Story Kit.
