html-diagram is a specialist skill for diagrams whose layout and notation clarify relationships, order, topology, state or quantities. The agent first names the question the reader should answer, then selects a diagram form and a rendering method among HTML and CSS, SVG, Canvas and WebGL.
It adds interaction such as sequencing, filtering, path tracing, pan and zoom only when that helps, and delivers one HTML file that works without a build step. It is meant to be invoked directly or routed from the broad html skill.
What You Get
- One self-contained HTML file with essential CSS and JavaScript inline
- A diagram form matched to the question: topology, sequence, process, state, hierarchy, timeline, matrix or quantitative view
- Keyboard-accessible controls and a reduced-motion fallback for any interaction
- A short handoff naming the form chosen and the assumptions made
Common Use Cases
- Showing how services, queues and stores connect in a system
- Tracing a request through a sequence of messages over time
- Mapping the states and transitions of an order or a workflow
- Laying out ownership and containment in an organization or a codebase
What Makes It Unique
It does not force every topic into one notation. The agent picks the grammar and the medium from the information, uses stable node positions so states can be compared, and verifies label collisions, edge routing, clipped nodes and keyboard operation at wide and narrow widths.
Under The Hood
The skill is a single instruction file with no scripts. The agent writes one HTML file to your workspace and, when a design-artifact skill is installed, reads it for visual direction. Verification means opening the page in a browser at wide and narrow widths and exercising its states. The instructions do not fetch web content or call external services, and they require none for the finished diagram.