Claude Code plugin and portable Agent Skill · Versioned and extensively documented, with repository-described cross-platform CI and validation gates; adoption and activity metadata are uncertain because supplied dates are future-dated.

cathrynlavery/diagram-design

MIT-licensed Claude Code plugin and portable Agent Skill for producing 39 opinionated editorial diagram types as self-contained HTML with inline SVG, importing draw.io and Mermaid structure for redrawing, applying brand profiles, optionally adding accessible motion, and exporting diagram-only SVG or PNG through documented supporting workflows.

diagrammingdata visualizationtechnical communicationarchitecture documentationproduct designeditorial designaccessibilitybrand systems
Routing score
80.0
Readiness
Usable for diagram generation with strong repository-authored checks, but output quality, host compatibility, and security claims were not independently verified from executed tests.
License
MIT
Maintenance
active
Components
16
Revision
0

Selection

Select when

  • You want Claude Code to produce polished standalone HTML and SVG diagrams.
  • You need one of the repository's 39 documented diagram types.
  • You want a consistent no-shadow editorial style with strict complexity budgets.
  • You need to redraw draw.io or Mermaid content rather than preserve its original visual layout.
  • You need reusable brand tokens or separate client profiles.
  • You need accessible SVG names and reduced-motion behavior by default.
  • You want format, frame size, detail, and audience controls for imported diagrams.
  • You value local validation scripts and repository-described CI checks for generated assets.
  • You need the same diagram skill across several Agent Skills-compatible hosts.

Boundaries

Avoid when

  • A paragraph, list, or table would communicate the material more clearly.
  • You need a collaborative visual canvas or direct Figma editing.
  • You require exact preservation of draw.io or Mermaid coordinates and styling.
  • You need an unsupported diagram grammar or more density than its budgets allow.
  • You cannot accept an opinionated editorial style or its strict connector and spacing rules.
  • You need guaranteed offline typography because default templates may reference Google Fonts.
  • You need PNG export but cannot install or run the documented Playwright and Chromium dependency.
  • You require independently audited parser security or verified production compatibility.
  • You need maintained security fixes for pinned older commits or forks.

Strengths

Capabilities

Packages a native Claude Code plugin around a shared SKILL.md workflow.Provides 39 stated visual types spanning architecture, process, hierarchy, quantitative charts, planning, data platforms, UML, journeys, deployment, and schemas.Generates self-contained HTML with inline SVG and embedded CSS as the default deliverable.Routes behavior-rich requests through seven semantic patterns before selecting the nearest visual layout.Ships minimal-light, minimal-dark, and full-editorial static variants for every stated visual type.Keeps static output as the default and offers optional reveal, step, and loop motion with reduced-motion fallbacks.Onboards brand colors and typography from a website, local design source, installed skill, or manually supplied tokens.Stores reusable client profiles locally and resolves project-specific profiles through a marker file.Extracts structural information from common draw.io containers without rendering their source content.Extracts Mermaid text from dedicated files or fenced Markdown blocks without executing or following source directives.Redraws imported structure according to format, size, detail, and audience controls rather than preserving source layout.Reports a fidelity ledger describing imported content that was merged, collapsed, or dropped.Exports diagram-only SVG and PNG through documented workflows, with Playwright required for PNG rasterization.Provides plugin commands for export, draw.io import, Mermaid import, profile management, and environment diagnostics.Ships a local self-check script for generated output safety, accessibility, and basic motion conformance.Defines accessible SVG naming through role, aria-labelledby, title, description, and collision-resistant IDs.Enforces opinionated density, accent, connector, typography, spacing, and complexity constraints in the skill guidance.Documents extensive repository validation for package structure, accessibility, geometry, rendering, quantitative encodings, imports, motion, screenshots, and documentation sync.States that GitHub Actions validates pull requests and pushes on Linux, Windows, and macOS.Supports the same skill files across Claude Code, Codex, Factory Droid, Pi, and other Agent Skills-compatible tools.

Risk profile

Risks and limitations

  • Uncertainty: only selected first-party files were supplied, so referenced scripts, templates, examples, manifests, workflows, tests, and generated assets were not directly inspected.
  • Uncertainty: supplied repository creation, update, and push dates are in 2026, so recency, popularity, and development activity cannot be reconciled with the analysis context.
  • Uncertainty: metadata describes 38 diagram types while the README, skill, and plugin manifests describe 39.
  • Uncertainty: the README says there is no build step or JavaScript for static variants, but optional motion and export workflows introduce JavaScript or browser automation in selected paths.
  • Repository-authored CI, parser-safety, rendering, accessibility, and compatibility claims were not independently executed or audited.
  • The strict nine-node default and other type budgets may require splitting dense production diagrams into overview and detail views.
  • Brand onboarding fetches public website content and may load public font URLs, so network access and external service behavior affect the workflow.
  • Default HTML may request Google Fonts, so it is not strictly network-free unless fonts are changed or unavailable fallbacks are accepted.
  • PNG export requires Python Playwright and Chromium according to the supplied documentation.
  • Managed updates may replace direct edits to the installed style guide, although local profiles and project markers are documented to persist.
  • The security policy supports only the latest main-branch commit and says older commits and forks are not separate support channels.
  • Visual clarity, brand fidelity, quantitative correctness, and accessibility still require project-specific human review despite automated checks.

First-party files say draw.io and Mermaid inputs are parsed as untrusted text without rendering, fetching, executing, or following click targets; motion linting rejects unapproved scripts, remote assets, imports, executable attributes, and unsafe URLs. Static output is normally script-free, a packaged self-check is provided, and vulnerabilities can be reported privately through GitHub. These controls are repository claims and were not independently audited; supported security fixes apply only to the latest main commit.

Component inventory

16 documented components

Claude Code plugin

Diagram Design for Claude Code

Native plugin packaging the shared skill and five namespaced slash commands; extends Claude Code through context injection, command surfaces, and guided workflows.

.claude-plugin/plugin.json
Agent Skill

diagram-design

Creates branded editorial diagrams in self-contained HTML with inline SVG/CSS, choosing among 39 visual types and optional semantic patterns.

skills/diagram-design/SKILL.md
Claude Code command

Export Diagram

Exports a generated HTML diagram’s SVG as standalone SVG, PNG, or both; PNG uses Playwright.

commands/export-diagram.md
Claude Code command

Import draw.io

Extracts structure from supported draw.io containers and redraws it in the editorial design system with selected output dials.

commands/import-drawio.md
Claude Code command

Import Mermaid

Parses Mermaid text or fenced Markdown blocks without rendering or executing them, then redraws the content editorially.

commands/import-mermaid.md
Claude Code command

Client Profiles

Manages named brand profiles shared across hosts and resolves project selection through a .diagram-design marker.

commands/profile.md
Claude Code command

Doctor

Runs documented environment diagnostics for the packaged plugin and its routed support files.

commands/doctor.md
Workflow

Brand Onboarding

Extracts palette and typography from a website, installed skill, local design-system folder, or pasted tokens, checks contrast, and proposes a style-guide update.

skills/diagram-design/references/onboarding.md
Workflow module

Semantic Pattern Routing

Selects one of seven behavior patterns before choosing the nearest visual layout for queues, policy traces, trust boundaries, governance, and related cases.

skills/diagram-design/references/semantic-patterns.md
Workflow module

Accessible Motion

Adds optional none, reveal, step, or loop presentation modes while preserving a complete static, reduced-motion, print, and export state.

skills/diagram-design/references/animation.md
Web gallery

Diagram Gallery

Browsable gallery of all 39 diagram types with light, dark, and full-editorial variants.

skills/diagram-design/assets/index.html
Template module

Diagram Templates

Static light, dark, full-editorial, terminal, and optional motion scaffolds for creating self-contained diagrams.

skills/diagram-design/assets/
Validation utility

Packaged Output Self-Check

Checks generated diagrams for the accessible-SVG contract, single-file safety, and basic motion requirements.

skills/diagram-design/scripts/self_check.py
Pi extension prompts

Pi Prompt Surface

Exposes export, Mermaid import, profile, and doctor prompt templates, plus explicit skill invocation in Pi.

prompts/
Codex plugin

Codex Plugin

Packages the same shared diagram skill for Codex with read/write capabilities and marketplace metadata.

.codex-plugin/plugin.json
Factory Droid plugin

Factory Droid Plugin

Packages the shared diagram skill and plugin command surface for Factory Droid.

.factory-plugin/plugin.json

Technical profile

Requirements and configuration

License
MIT
RepositoryLanguage
HTML
PluginVersion
2.6.7 in supplied Claude, Codex, and Factory manifests
SkillMetadataVersion
2.6
PrimaryArtifact
skills/diagram-design/SKILL.md
DefaultOutput
Self-contained HTML with inline SVG and CSS
OptionalOutputs
Diagram-only SVG and PNG
StaticVariants
Minimal light, minimal dark, and full editorial
DiagramTypes
39 according to README, skill, and manifests; metadata description says 38
SemanticPatterns
Seven behavior patterns routed independently of visual type
ImportFormats
draw.io, draw.io XML, PNG-embedded draw.io, SVG-embedded draw.io, Mermaid files, and fenced Mermaid blocks in Markdown
ProfileStorage
~/.diagram-design/profiles/ with project selection through a .diagram-design marker
CoreRuntime
Agent host plus a modern browser for viewing generated HTML
ScriptRuntime
Python for bundled extractors and self-checks
PngDependency
Playwright and Chromium
NetworkBehavior
Static files are self-contained except optional Google Fonts; onboarding may fetch a website
Validation
Repository describes GitHub Actions and source, render, geometry, accessibility, motion, import, packaging, quantitative, and documentation gates
SecuritySupport
Latest commit on main only
Installation
Native Claude Code plugin marketplace or shared-skill linking/copying paths documented for supported hosts

Classification

How it enters the stack

Context InjectionTool Surface

Evidence: a native Claude plugin loads a routed SKILL.md and commands; bundled Python extractors and self-checks add local executable helpers. Basic Claude installation is short, while optional PNG export and development checks require extra dependencies.

Claude Extension · low setup effort · high confidence · automated

Evidence and risk

Primary sources

first_party_fileREADME.mdhttps://github.com/cathrynlavery/diagram-design/blob/main/README.md
first_party_fileSECURITY.mdhttps://github.com/cathrynlavery/diagram-design/blob/main/SECURITY.md
first_party_fileCONTRIBUTING.mdhttps://github.com/cathrynlavery/diagram-design/blob/main/CONTRIBUTING.md

Routing context

Conflicts, complements, and synergies

extends

gh_anthropics_claude_code

The repository packages a native Claude Code plugin with a shared Agent Skill and plugin commands.

high confidence
overlaps

gh_owl_listener_designer_skills

Both extend Claude Code with design guidance, but this repository specializes in generated editorial diagrams rather than broad design practice.

medium confidence
overlaps

gh_mengto_skills

Both provide portable design-oriented Agent Skills, while this repository is focused on diagram taxonomy, imports, and SVG output.

medium confidence
complements

gh_conardli_garden_skills

Garden Skills covers broader visual and editorial artifact workflows that can use diagrams produced by this specialized skill.

medium confidence
complements

gh_codeswithroh_tastemaker

Tastemaker focuses on reference-grounded UI taste, while Diagram Design provides branded diagram generation and source redrawing.

medium confidence
complements

gh_jakubkrehel_skills

The interface-review collection can review surrounding UI, while this repository generates standalone diagram artifacts.

medium confidence