Give your AI agent project-aware knowledge of Dashfy — scaffolding, extensions, widgets, and dashboard layout.
Skills give AI assistants like Cursor, Claude Code, and Codex project-aware context about Dashfy. When installed, your agent knows how to scaffold apps, find and install extensions, set up widgets and data sources, and create dashboards using the correct CLI, registry, and MCP APIs for your project.
For example, you can ask your AI assistant to:
- "Add the GitHub extension and show my repo's pull requests."
- "Scaffold a new dashboard app with the system monitor extension."
- "Create a dashboard with CPU, memory, and disk widgets."
- "Find an extension for my JSON API and set it up."
The skill reads your project (dashfy.json, dashfy.config.yml, installed
@dashfy/ext-* packages) and provides the assistant with the detected paths,
configured registries, and available widgets so it can generate correct code on the
first try.
Install
Add the Dashfy skill to your AI agent with the CLI:
pnpm dlx dashfy@latest add skillThe Cursor plugin bundles the skill together with the project rules and the MCP server in a single installable package.
What's included
The skill provides your AI assistant with the following knowledge:
Project context
The skill runs dashfy info --json to read your project configuration: detected
files, dashfy.json paths, configured registries, and installed @dashfy/ext-*
packages. This grounds every suggestion in your actual project.
CLI commands
Full reference for every command — init, add, remove, search, view, docs,
info, doctor, registry, and mcp — including flags and dry-run mode. See the
CLI reference.
Registry and addresses
How extensions are resolved from registries, how to classify an address (bare name,
@namespace/name, owner/repo/name, URL, or local file), and how to author and host
custom registries. See Registry.
Dashboard layout
The widget grid rules — every widget needs extension, widget, x, y, columns,
and rows; widgets stay inside the grid and must not overlap; only registered widgets
may be referenced. See Configuration.
MCP tools
The tools exposed by the dashfy mcp server so agents can discover and install
extensions programmatically. See the MCP server.
How it works
Project detection
The skill activates when it finds a dashfy.json or dashfy.config.yml in your
project.
Context injection
It runs dashfy info --json and injects the result into the assistant's context, so
suggestions reflect your detected files, registries, and installed extensions.
Pattern enforcement
The assistant follows Dashfy conventions: it uses dashfy add / dashfy remove
codemods instead of manual setup, keeps widgets within the dashboard grid, and
classifies registry addresses correctly (bare names resolve to @dashfy).
Extension discovery
Before generating code, the assistant uses dashfy search, dashfy docs, or the MCP
tools to find extensions and read their setup documentation.
Verification
After adding an extension, the assistant runs dashfy doctor to confirm the package,
widget registration, server API, environment variables, and starter block are all in
place.
Using with other agents
The skill is a standard agent skill (a SKILL.md with name and description), so
it works beyond Cursor:
- Cursor — install the Cursor plugin, which bundles the skill, project rules, and the MCP server.
- Claude Code, Codex, OpenCode, and others — use the skill directly, and connect
the MCP tools with
dashfy mcp init(see the MCP server reference for the supported clients).
Learn more
- CLI — every command, flag, and example.
- MCP server — tools for discovering and installing extensions.
- Registry — authoring and hosting extension registries.
- Cursor plugin — the skill, rules, and MCP in one package.
- Configuration — the dashboard configuration model.