Guide · Agent skill

How to use the Fireworks Tech Graph agent skill

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

Fireworks Tech Graph is an agent skill that lets an AI agent turn a text description of a system into a technical diagram, such as an architecture, UML, cloud, agent or workflow diagram. This guide is for people who have not used agent skills before. By the end you will know how to install the skill, how to ask for a diagram, and how to export the result as PNG, offline HTML or an animated GIF.

Fireworks Tech Graphyizhiyanhua-ai/fireworks-tech-graph · MIT Download the skill

Quick start

1Download the ZIP. Get the skill ZIP from the download link on our page.
2Unzip into the skills folder. Unzip into ~/.claude/skills/ for all projects, or into .claude/skills/ inside one project.
3Or upload in Claude apps. In the Claude apps, go to Settings, then Capabilities, then Skills, and upload the skill.
4Confirm the scripts run. Run the version command with Python 3 to check the installed copy works.
5Describe your system. Ask the agent for a diagram with components, connections and diagram type named in plain language.
6Check and export. Let the agent validate, render, check and export PNG, then review the image at reading size.
7Optional: HTML or GIF. Ask for an offline HTML page or an animated GIF if the diagram style supports it.

What an agent skill is and what this one does

An agent skill is a folder of instructions and optional scripts that an AI agent loads when a task matches the skill's description. This one is named fireworks-tech-graph and is written for Codex and Claude Code.

Once installed, the agent can create precise SVG technical diagrams from a brief. SVG is the main output, and the skill also supports exporting PNG, a self-contained offline HTML page with pan and zoom, and animated GIF for supported diagram styles.

Diagram types named in the skill: architecture, UML, agent, cloud and workflow diagrams.
Not for photos, raster art or statistical charts; the skill description excludes these.
The agent checks labels, connectors and layout with bundled scripts rather than trusting a successful render alone.

Install the skill from the ZIP or the Claude apps

Download the ZIP of the skill from the download link on our page. Then choose one of the following methods.

For Claude Code, unzip the folder into ~/.claude/skills/ to make it available everywhere, or into .claude/skills/ inside a specific project to limit it to that project.

In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the skill from the ZIP.

Personal install: unzip into ~/.claude/skills/.
Project install: unzip into .claude/skills/ in your project folder.
Claude apps: Settings → Capabilities → Skills, then upload.

Check that the scripts run

The skill's main script is scripts/fireworks.py and needs Python 3. The skill's own instructions tell the agent to resolve the skill folder as SKILL_ROOT, using the CLAUDE_SKILL_DIR variable in Claude Code. You can run the version command to see which version you have installed. Run doctor only when something is missing, such as a renderer or font.

SKILL_ROOT="${CLAUDE_SKILL_DIR:-/absolute/path/from-codex-skill-metadata}"
python3 "$SKILL_ROOT/scripts/fireworks.py" version

Ask for a diagram in plain language

You do not need to type commands yourself. Describe the system, the components, how they connect and the diagram type, then ask for the output. Words such as 'draw', 'diagram' or 'architecture diagram' are the kind of request the skill is written for.

Good briefs name the parts and the connections. If a fact about the system is missing, such as which protocol a link uses, say so or leave it out. The skill is told not to invent responsibilities, protocols or metrics, so the diagram is only as accurate as your brief.

If you do not name a style, the skill defaults to Style 1 Flat Icon. Styles 9 to 12 default to C4, cloud, event and observability contracts.

Example: 'Draw an architecture diagram of a web app with a load balancer, two API servers and a Postgres database.'
Example: 'Make a UML sequence-style diagram of a user logging in through an OAuth provider.'
Example: 'Create a workflow diagram of our nightly data pipeline and export a PNG for the README.'

Generate, check and export the files

The agent normally validates the input, renders the SVG, runs a check and exports a PNG. The check looks at SVG identity, arrow references, overlapping elements and layout. A successful render is not treated as proof of visual quality, so the agent is also asked to review the PNG at reading size when image viewing is available.

To request HTML, ask for an offline HTML version. To request an animated GIF, use wording such as 'Generate a GIF' or 'Animate this diagram'. GIF motion is enabled for documented scene contracts in styles 1 to 12, and the skill does not promise arbitrary same-style layouts.

SVG: the canonical, editable source file.
PNG: exported from the SVG with a set width, default 1920 in the example command.
Offline HTML: one sanitized file with pan and zoom, source copy and static image downloads.
GIF: default 960px, 20fps, 5.75s duration, with a .motion.json report.

python3 "$SKILL_ROOT/scripts/fireworks.py" validate architecture input.json
python3 "$SKILL_ROOT/scripts/fireworks.py" render architecture input.json diagram.svg --report layout.json
python3 "$SKILL_ROOT/scripts/fireworks.py" check diagram.svg
python3 "$SKILL_ROOT/scripts/fireworks.py" export-png diagram.svg diagram.png --width 1920

Know what the skill needs and what it runs

The skill runs local scripts. The source states that it uses no network access. PNG export needs renderer and font dependencies; the doctor command checks for them, but the exact requirements are not confirmed in the provided files. The project's contributing notes list CairoSVG or rsvg-convert, and Node.js 22.12 or later for optional Chromium export.

Review scripts before running them on sensitive projects. The agent runs local Python scripts, so treat it like any tool that executes code on your machine. Publishing or sending a diagram somewhere remote needs its own explicit approval; the skill does not treat earlier approval as permission for that.

Runs scripts: yes. Python 3 is required for scripts/fireworks.py.
Uses network: no, according to the skill's listing.
Do not put secrets, tokens or confidential internal details into the brief if you do not want them in the diagram.

What people use it for

Drafting an architecture diagram for an engineering design review.
Producing an SVG and PNG figure for a project README.
Creating a workflow diagram for a slide deck, with the SVG kept as editable source.
Turning a written system description into an offline HTML page that readers can pan and zoom.

Frequently asked questions

Is the Fireworks Tech Graph skill free to use?

The skill is published under the MIT licence, as stated on the project page. Check the LICENSE file in the repository for the full terms before redistributing it.

What do I need to run the Fireworks Tech Graph skill?

You need Python 3 to run scripts/fireworks.py. PNG export needs renderer and font dependencies, which the doctor command can check, though the exact list is not confirmed in the provided files.

How do I install the fireworks-tech-graph skill in Claude Code?

Download the ZIP from our page and unzip it into ~/.claude/skills/ for all projects, or into .claude/skills/ inside a project. In the Claude apps you can instead upload it under Settings, Capabilities, Skills.

Can the skill make photos, raster art or bar charts?

No. The skill is for technical diagrams, and its description explicitly excludes photos, raster art and statistical charts.

Does the skill need internet access?

The skill's listing says it does not use the network. It does run local scripts, so review them if you work with sensitive material.

Why does the diagram have a wrong connection or missing detail?

The output is only as accurate as your brief, and the skill is designed not to invent responsibilities, protocols or metrics. Give exact component names and connections, and review the PNG before using it.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits engineers and documentation writers who want clean SVG and PNG technical diagrams generated from a text brief, and who want to keep the SVG as editable source. It is a poor fit for photos, raster art, statistical charts, or for diagrams of systems whose facts you have not yet confirmed.

Download the skill