FlowForge logo

FlowForge

Community
wentong2022-arch
FlowForge

TRIGGER when: user asks to draw, illustrate, or visualize any process, structure, or concept — "画个流程图", "draw an architecture diagram", "帮我画个对比图", "visualize this process", "make a diagram for this doc", or mentions 流程图/架构图/示意图/对比图/时序图/泳道图/时间线. Also trigger when user provides a document/article and asks for illustrations or diagrams. DO NOT trigger for: chart/plot requests (bar charts, line graphs — those need a charting library, not draw.io), or artistic image requests (logos, mascots, drawings of objects/people — those need an image tool).

Overview

Publisherwentong2022-arch
Repositoryflowforge-skill
Skill nameFlowForge
Stars
139
Forks
16
Bundled files
12
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 12 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by wentong2022-arch on GitHub. Read the source before you install it.

Installation

Install the FlowForge AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/wentong2022-arch/flowforge-skill/tree/main/skills/FlowForge
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/wentong2022-arch/flowforge-skill.git /tmp/flowforge-skill
mkdir -p .claude/skills
cp -r /tmp/flowforge-skill/skills/FlowForge .claude/skills/FlowForge
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable FlowForge in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads FlowForge on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so FlowForge is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

FlowForge — Draw.io Diagram Skill

Generate professional diagrams as draw.io XML files. Supports 13 diagram types: flows, comparisons, layer stacks, cycles, trees, hub-and-spoke, columns, matrices, funnels, timelines, sequence diagrams, and swimlanes.

Usage

bash
/FlowForge "OAuth 2.0 authorization code flow"
/FlowForge path/to/design-doc.md
/FlowForge "CI/CD pipeline" --type loop
/FlowForge "Monolith vs Microservices" --type compare --theme morandi
/FlowForge "System Architecture" --lang en

Workflow

Step 1: Understand & Propose (one confirmation, not three)

If given a document: read it fully, identify 1–3 places that benefit from a diagram, and for each note the insert position, the topic (one sentence), and the recommended type (see Diagram Type Reference).

If given a concept directly: identify the core message and pick the type via the Selection Guide.

Then present one combined proposal and ask for confirmation once. The proposal contains:

  • Diagram type and theme (default: tech-blue; other options: morandi, mint, terracotta, indigo)
  • An ASCII sketch: all nodes with labels, arrow directions and labels, grouping, color-role tags, layout direction

Example sketch:

Direction: left-to-right | Nodes: 4 | Type: flow | Theme: tech-blue

[primary] Code Commit  →  [process] Build  →  [process] Test  →  [accent] Deploy
                                                    ↓ (fail)
                                              [error] Alert

Skip the confirmation entirely when the request already pins everything down (user gave --type/--theme or an unambiguous description), or when no human can respond (subagent, batch, CI). In that case proceed with defaults and state your assumptions in the final report.

Step 2: Generate draw.io XML

  1. Read references/layouts.md — the layout algorithm for the chosen type — and compute all coordinates from its formulas
  2. Apply theme colors from themes.md
  3. Use element templates from xml-reference.md
  4. For an unfamiliar type, read the matching example file listed in examples.md

Step 3: Validate (mandatory)

bash
python3 {skill_dir}/scripts/validate.py <file.drawio> --theme <theme>

This deterministically checks: XML well-formedness, unique ids, dangling arrow references, orthogonal edge style, canvas bounds, node overlap, z-order (containers before children), font sizes, text-fit estimates, and theme color compliance. Fix every ERROR and re-run until clean. WARNs are heuristics — judge each one, fix the real ones.

Step 4: Visual self-check (when draw.io CLI is available)

bash
bash {skill_dir}/scripts/render.sh <file.drawio>   # writes a .png next to the file

If the script reports draw.io is not installed, skip this step (the user can enable it once with brew install --cask drawio). Otherwise Read the PNG and inspect it with fresh eyes: text overflowing nodes, crowded regions, arrows colliding with nodes, color rhythm, overall balance. Fix and re-render until it looks right — this catches what rules can't.

Step 5: Save & Deliver

Save the .drawio file to the output directory (see Configuration). After saving, report:

Iteration

If the user wants changes after viewing:

  1. Read the current .drawio file
  2. Modify the XML per feedback
  3. Save as {name}_v2.drawio (preserve previous version), then re-run Steps 3–4

Configuration

Optional per-project config file ./.flowforge.json (in the working directory, never inside the skill directory — the skill directory may be read-only or shared across projects):

json
{
  "defaultTheme": "tech-blue",
  "outputDir": "./diagrams",
  "language": "auto"
}

If the file is absent, use exactly these defaults — do not interview the user. "language": "auto" means match the user's language; technical abbreviations (API, LLM, CI/CD) stay in English regardless. Per-invocation flags (--theme, --lang) override config.


Diagram Type Reference

Core Types

TypeCodeBest ForExample
Linear FlowflowSequential steps A→B→CAPI call flow, data pipeline
Vertical Flowflow-verticalTop-down processesDecision process, approval chain
ComparisoncompareA vs B side by sideTraditional vs AI approach
Layer StacklayersTech stack, tiersSystem architecture layers
CycleloopIterative processesML training loop, CI/CD
TreetreeHierarchies, taxonomiesOrg chart, decision tree
Hub & SpokehubCore concept + branchesProduct feature map

Extended Types

TypeCodeBest For
Parallel Columnscolumns3+ parallel concepts
MatrixmatrixMulti-dimension comparison
FunnelfunnelFiltering, conversion
TimelinetimelineVersion evolution, history
SequencesequenceComponent interactions
SwimlaneswimlaneCross-role processes (who does what)

Selection Guide

Content PatternRecommended Type
Sequential stepsflow or flow-vertical
Two things comparedcompare
3+ parallel conceptscolumns or hub
Layered systemlayers
Iterative/cyclical processloop
One core, many branches (incl. mind-map style)hub or tree
Components communicatingsequence
Change over timetimeline
Multi-criteria evaluationmatrix
Progressive filteringfunnel
Steps owned by different roles/systemsswimlane

Color Usage Principles

This is the most important section for visual quality.

The 4+6 Rule

Each theme has 4 base colors (primary, process, accent, neutral) and 6 extended colors (success, warning, error, secondary, storage, group). The primary family should always dominate; the amount of non-primary accents scales with diagram size.

Color Budget by Diagram Size

Node CountTarget ColorsGuideline
3-5 nodes2-3 colorsPrimary family dominates. At most 1 accent node.
6-8 nodes3-4 colorsPrimary ~60%. Add 1-2 non-primary at semantic turning points.
9+ nodes4-5 colorsPrimary ~50%. Distribute 2-3 non-primary for visual rhythm.

6+ colors almost always looks bad regardless of diagram size.

Color Assignment Strategy

  1. Primary dominates — primary and process cover the largest visual area and the biggest nodes.
  2. Accent is a scalpel, not a paintbrush — at most 1-2 key highlights, never the largest node.
  3. Scale with length — a 4-node flow with 2 colors looks clean; a 9-node flow with 2 colors is a monotonous wall. Add non-primary colors at natural semantic boundaries to create rhythm.
  4. 3-consecutive guideline — if you spot 4+ adjacent same-color nodes, promote the most semantically distinctive one.
  5. Distribute spatially — non-primary nodes spread across the diagram, not clustered in one region.

Semantic Color Heuristics

Node TypeRecommended Color
User input / starting pointprimary
Standard processing stepprocess
Decision / branching pointwarning or accent
Domain-specific / translation stepsecondary
Key transformation / highlightaccent
Data source / external systemneutral
Error path / fallbackaccent or error
Successful output / endpointsuccess or primary
Database / storagestorage

Good/bad worked examples of color rhythm are at the bottom of themes.md.


Design Principles

  1. Narrative titles — Use "How OAuth 2.0 Authorizes a User" not "OAuth 2.0 Diagram".
  2. Bilingual labels — Use the user's language. Technical abbreviations (API, LLM, CI/CD) stay in English regardless.
  3. Orthogonal flow arrows — Every arrow between nodes includes edgeStyle=orthogonalEdgeStyle; so it runs horizontal/vertical only. Exception: decorative connectors with endArrow=none (hub rays, timeline axis/ticks) may run straight.

Gotchas — Common Failure Points

The validator (Step 3) catches most geometry/reference mistakes after the fact, but knowing these saves a fix cycle:

  1. Rich node title duplication — the title lives in a separate text cell; keep value="" on the background rect or it renders twice.
  2. Arrow ignores exit/entry points — always specify exitX/exitY/entryX/entryY; otherwise draw.io auto-routes, usually badly.
  3. HTML entities in value attributes — value is HTML inside an XML attribute: &lt; &gt; &amp;, and &amp;bull; for bullets. One missing escape breaks the whole file (validator: parse error).
  4. Filled boxes drawn after their children hide them — define lane bands and any filled container BEFORE the nodes inside them (validator: z-order check).
  5. Canvas too small — compute pageWidth/pageHeight from the layout formulas: max(x + w) + CANVAS_PAD (validator: bounds check).
  6. Forgetting parent="1" — every content element needs it unless intentionally grouped (validator: parent check).
  7. Diamond text overflow — rhombus shapes have ~60% usable width; keep labels to 2-3 short words, font 11.
  8. Snake layout wrap arrows — the wrap-around arrow exits the BOTTOM of the last node in a row and enters the TOP of the next row's first node, not sideways.
  9. Cycle return arrow — in loop diagrams, route the last→first arrow around the outside of the rectangle, never through the center.
  10. Off-by-one in widths — n nodes have n-1 gaps: canvas width = PAD×2 + n×NODE_W + (n-1)×GAP_H (validator: bounds/overlap).
  11. Branch/merge arrows go diagonal — when exit/entry points aren't aligned, only edgeStyle=orthogonalEdgeStyle; keeps them orthogonal — include it in every flow arrow (validator: style check).

Supporting Files

Read them when the workflow step calls for them — don't load everything upfront.

FileWhen to read
references/layouts.mdStep 2, before computing coordinates — global constants + all 13 layout algorithms
themes.mdStep 2, when applying colors — 5 themes × 10 semantic colors + 7 text colors, color-rhythm examples
xml-reference.mdStep 2, when writing XML — canvas boilerplate, element templates, arrow direction table
examples.mdStep 2 — coordinate-calculation walkthroughs + index of example files
scripts/validate.pyStep 3 — run on every generated file (don't read it, run it)
scripts/render.shStep 4 — export PNG for the visual self-check

Example .drawio files (in examples/)

Real, working files to read as reference when generating a similar type:

  • examples/flow-cicd.drawio — Linear flow, 4 nodes, tech-blue
  • examples/compare-monolith-vs-micro.drawio — Comparison, 2 columns × 3 rows, morandi
  • examples/loop-ml-training.drawio — Cycle, 4 nodes clockwise, mint
  • examples/tree-storage-decision.drawio — Decision tree with diamonds, mint
  • examples/hub-agent-capabilities.drawio — Hub & spoke, 6 spokes with rays, indigo
  • examples/timeline-llm-history.drawio — Timeline with axis + ticks, terracotta

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the FlowForge AI skill do?

TRIGGER when: user asks to draw, illustrate, or visualize any process, structure, or concept — "画个流程图", "draw an architecture diagram", "帮我画个对比图", "visualize this process", "make a diagram for this doc", or mentions 流程图/架构图/示意图/对比图/时序图/泳道图/时间线. Also trigger when user provides a document/article and asks for illustrations or diagrams. DO NOT trigger for: chart/plot requests (bar charts, line graphs — those need a charting library, not draw.io), or artistic image requests (logos, mascots, drawings of objects/people — those need an image tool).

Why use FlowForge on TypingMind?

Because you install it once and use it with any model. FlowForge is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install FlowForge in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/wentong2022-arch/flowforge-skill/tree/main/skills/FlowForge. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use FlowForge?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with FlowForge?

As many as you like. As long as a model supports skills, you can use FlowForge with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the FlowForge AI skill free?

Yes. It is published on GitHub by wentong2022-arch under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇