Back to directory

fireworks-tech-graph

An AI agent skill, usable unchanged in Codex and Claude Code, that turns natural-language descriptions into geometry-checked SVG technical diagrams, PNG exports, SVG-to-GIF motion and interactive HTML.

Visit website (opens in a new tab)yizhiyanhua-ai.github.io
fireworks-tech-graph

fireworks-tech-graph is an AI agent skill that works unchanged in both Codex and Claude Code, turning a natural-language description of a system into a polished, geometry-checked technical diagram. From one request, it can produce an SVG diagram, a high-resolution PNG export, a validated SVG-to-GIF animation for a focused part of the diagram, and an offline interactive HTML version.

It ships with 11 generator-backed visual styles and one AI-authored style, and four additional engineering-focused styles add executable contracts specifically for C4 architecture reviews, cloud-deployment diagrams, event-stream diagrams and reliability investigations, alongside dedicated patterns for AI and agent workflow diagrams and support for all 14 standard UML diagram types.

A companion command-line script reports the installed package version and current Git state, distinguishes SVG and HTML support from the optional dependencies needed for PNG or GIF export, and exports an SVG diagram to a PNG of a chosen width, checking and writing the output atomically.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.