markdown-mermaid-writing

Markdown and Mermaid Writing

Overview

This skill teaches you — and enforces a standard for — creating scientific documentation using markdown with embedded Mermaid diagrams as the default and canonical format.

The core bet: a relationship expressed as a Mermaid diagram inside a .md file is more valuable than any image. It is text, so it diffs cleanly in git. It requires no build step. It renders natively on GitHub, GitLab, Notion, VS Code, and any markdown viewer. It uses fewer tokens than a prose description of the same relationship. And it can always be converted to a polished image later — but the text version remains the source of truth.

"The more you get your reports and files in .md in just regular text, which mermaid is as well as being a simple 'script language'. This just helps with any downstream rendering and especially AI generated images (using mermaid instead of just long form text to describe relationships < tokens). Additionally mermaid can render along with markdown for easy use almost anywhere by humans or AI."

— Clayton Young (@borealBytes), 2026-02-19

When to Use This Skill

Use this skill when:

• Creating any scientific document — reports, analyses, manuscripts, methods sections
• Writing any documentation — READMEs, how-tos, decision records, project docs
• Producing any diagram — workflows, data pipelines, architectures, timelines, relationships
• Generating any output that will be version-controlled — if it's going into git, it should be markdown
• Working with any other skill — this skill defines the documentation layer that wraps every other output
• Someone asks you to "add a diagram" or "visualize the relationship" — Mermaid first, always

Do NOT start with Python matplotlib, seaborn, or AI image generation for structural or relational diagrams. Those are Phase 2 and Phase 3 — only used when Mermaid cannot express what's needed (e.g., scatter plots with real data, photorealistic images).

🎨 The Source Format Philosophy

Why text-based diagrams win

What matters	Mermaid in Markdown	Python / AI Image
Git diff readable	✅	❌ binary blob
Editable without regenerating	✅	❌
Token efficient vs. prose	✅ smaller	❌ larger
Renders without a build step	✅	❌ needs hosting
Parseable by AI without vision	✅	❌
Works in GitHub / GitLab / Notion	✅	⚠️ if hosted
Accessible (screen readers)	✅ accTitle/accDescr	⚠️ needs alt text
Convertible to image later	✅ anytime	— already image

The three-phase workflow

flowchart LR
    accTitle: Three-Phase Documentation Workflow
    accDescr: Phase 1 Mermaid in markdown is always required and is the source of truth. Phases 2 and 3 are optional downstream conversions for polished output.

    p1["📄 Phase 1<br/>Mermaid in Markdown<br/>(ALWAYS — source of truth)"]
    p2["🐍 Phase 2<br/>Python Generated<br/>(optional — data charts)"]
    p3["🎨 Phase 3<br/>AI Generated Visuals<br/>(optional — polish)"]
    out["📊 Final Deliverable"]

    p1 --> out
    p1 -.->|"when needed"| p2
    p1 -.->|"when needed"| p3
    p2 --> out
    p3 --> out

    classDef required fill:#dbeafe,stroke:#2563eb,stroke-width:2px,color:#1e3a5f
    classDef optional fill:#fef9c3,stroke:#ca8a04,stroke-width:2px,color:#713f12
    classDef output fill:#dcfce7,stroke:#16a34a,stroke-width:2px,color:#14532d

    class p1 required
    class p2,p3 optional
    class out output

Phase 1 is mandatory. Even if you proceed to Phase 2 or 3, the Mermaid source stays committed.

What Mermaid can express

Mermaid covers 24 diagram types. Almost every scientific relationship fits one:

Use case	Diagram type	File
Experimental workflow / decision logic	Flowchart	references/diagrams/flowchart.md
Service interactions / API calls / messaging	Sequence	references/diagrams/sequence.md
Data model / schema	ER diagram	references/diagrams/er.md
State machine / lifecycle	State	references/diagrams/state.md
Project timeline / roadmap	Gantt	references/diagrams/gantt.md
Proportions / composition	Pie	references/diagrams/pie.md
System architecture (zoom levels)	C4	references/diagrams/c4.md
Concept hierarchy / brainstorm	Mindmap	references/diagrams/mindmap.md
Chronological events / history	Timeline	references/diagrams/timeline.md
Class hierarchy / type relationships	Class	references/diagrams/class.md
User journey / satisfaction map	User Journey	references/diagrams/user_journey.md
Two-axis comparison / prioritization	Quadrant	references/diagrams/quadrant.md
Requirements traceability	Requirement	references/diagrams/requirement.md
Flow magnitude / resource distribution	Sankey	references/diagrams/sankey.md
Numeric trends / bar + line charts	XY Chart	references/diagrams/xy_chart.md
Component layout / spatial arrangement	Block	references/diagrams/block.md
Work item status / task columns	Kanban	references/diagrams/kanban.md
Cloud infrastructure / service topology	Architecture	references/diagrams/architecture.md
Multi-dimensional comparison / skills radar	Radar	references/diagrams/radar.md
Hierarchical proportions / budget	Treemap	references/diagrams/treemap.md
Binary protocol / data format	Packet	references/diagrams/packet.md
Git branching / merge strategy	Git Graph	references/diagrams/git_graph.md
Code-style sequence (programming syntax)	ZenUML	references/diagrams/zenuml.md
Multi-diagram composition patterns	Complex Examples	references/diagrams/complex_examples.md

💡 Pick the right type, not the easy one. Don't default to flowcharts for everything. A timeline beats a flowchart for chronological events. A sequence beats a flowchart for service interactions. Scan the table and match.

---

🔧 Core workflow

Step 1: Identify the document type

Check if a template exists before writing from scratch:

Document type	Template
Pull request record	templates/pull_request.md
Issue / bug / feature request	templates/issue.md
Sprint / project board	templates/kanban.md
Architecture decision (ADR)	templates/decision_record.md
Presentation / briefing	templates/presentation.md
Research paper / analysis	templates/research_paper.md
Project documentation	templates/project_documentation.md
How-to / tutorial	templates/how_to_guide.md
Status report	templates/status_report.md

Step 2: Read the style guide

Before writing any .md file: read references/markdown_style_guide.md.

Key rules to internalize:

• One H1 per document — the title. Never more.
• Emoji on H2 headings only — one emoji per H2, none in H3/H4
• Cite everything — every external claim gets a footnote [^N] with full URL
• Bold sparingly — max 2-3 bold terms per paragraph, never full sentences
• Horizontal rule after every </details> — mandatory
• Tables over prose for comparisons, configurations, structured data
• Diagrams over walls of text — if it describes flow, structure, or relationships, add Mermaid

Step 3: Pick the diagram type and read its guide

Before creating any Mermaid diagram: read references/mermaid_style_guide.md.

Then open the specific type file (e.g., references/diagrams/flowchart.md) for the exemplar, tips, and copy-paste template.

Mandatory rules for every diagram:

accTitle: Short Name 3-8 Words
accDescr: One or two sentences explaining what this diagram shows.

• No %%{init} directives — breaks GitHub dark mode
• No inline style — use classDef only
• One emoji per node max — at the start of the label
• snake_case node IDs — match the label

Step 4: Write the document

Start from the template. Apply the markdown style guide. Place diagrams inline with related text — not in a separate "Figures" section.

Step 5: Commit as text

The .md file with embedded Mermaid is what gets committed. If you also generated a PNG or AI image, those are supplementary — the markdown is the source.

---

⚠️ Common pitfalls

Radar chart syntax (radar-beta)

WRONG:

radar
title Example
x-axis ["A", "B", "C"]
"Series" : [1, 2, 3]

CORRECT:

radar-beta
title Example
axis a["A"], b["B"], c["C"]
curve series["Series"]{1, 2, 3}
max 3

• Use radar-beta not radar (the bare keyword doesn't exist)
• Use axis to define dimensions, not x-axis
• Use curve to define data series, not quoted labels with colon
• No accTitle/accDescr — radar-beta doesn't support accessibility annotations; always add a descriptive italic paragraph above the diagram

XY Chart vs Radar confusion

Diagram	Keyword	Axis syntax	Data syntax
XY Chart (bars/lines)	xychart-beta	x-axis ["Label1", "Label2"]	bar [10, 20] or line [10, 20]
Radar (spider/web)	radar-beta	axis id["Label"]	curve id["Label"]{10, 20}

Forgetting accTitle/accDescr on supported types

Only some diagram types support accTitle/accDescr. For those that don't, always place a descriptive italic paragraph directly above the code block:

Radar chart comparing three methods across five performance dimensions. Note: Radar charts do not support accTitle/accDescr.

radar-beta
...

---

🔗 Integration with other skills

With scientific-schematics

scientific-schematics generates AI-powered publication-quality images (PNG). Use the Mermaid diagram as the brief for the schematic:

Workflow:
1. Create the concept as Mermaid in .md (this skill — Phase 1)
2. Describe the same concept to scientific-schematics for a polished PNG (Phase 3)
3. Commit both — the .md as source, the PNG as a supplementary figure

With scientific-writing

When scientific-writing produces a manuscript, all diagrams and structural figures should use this skill's standards. The writing skill handles prose and citations; this skill handles visual structure.

Workflow:
1. Use scientific-writing to draft the manuscript
2. For every figure that shows a workflow, architecture, or relationship:
   - Replace placeholder with a Mermaid diagram following this skill's guide
3. Use scientific-schematics only for figures that truly need photorealistic/complex rendering

With literature-review

Literature review produces summaries with lots of relationship data. Use this skill to:

• Create concept maps (Mindmap) of the literature landscape
• Show publication timelines (Timeline or Gantt)
• Compare methodologies (Quadrant or Radar)
• Diagram data flows described in papers (Sequence or Flowchart)

With any skill that produces output documents

Before finalizing any document from any skill, apply this skill's checklist:

• Does the document use a template? If so, did I start from the right one?
• Are all diagrams in Mermaid with accTitle + accDescr?
• No %%{init}, no inline style, only classDef?
• Are all external claims cited with [^N]?
• One H1, emoji on H2 only?
• Horizontal rules after every </details>?

---

📚 Reference index

Style guides

Guide	Path	Lines	What it covers
Markdown Style Guide	references/markdown_style_guide.md	~733	Headings, formatting, citations, tables, Mermaid integration, templates, quality checklist
Mermaid Style Guide	references/mermaid_style_guide.md	~458	Accessibility, emoji set, color classes, theme neutrality, type selection, complexity tiers

Diagram type guides (24 types)

Each file contains: production-quality exemplar, tips specific to that type, and a copy-paste template.

references/diagrams/ — architecture, block, c4, class, complex_examples, er, flowchart, gantt, git_graph, kanban, mindmap, packet, pie, quadrant, radar, requirement, sankey, sequence, state, timeline, treemap, user_journey, xy_chart, zenuml

Document templates (9 types)

templates/ — decision_record, how_to_guide, issue, kanban, presentation, project_documentation, pull_request, research_paper, status_report

Examples

assets/examples/example-research-report.md — a complete scientific research report demonstrating proper heading hierarchy, multiple diagram types (flowchart, sequence, gantt), tables, footnote citations, collapsible sections, and all style guide rules applied.

---

📝 Attribution

All style guides, diagram type guides, and document templates in this skill are ported from the SuperiorByteWorks-LLC/agent-project repository under the Apache-2.0 License.

• Source: https://github.com/SuperiorByteWorks-LLC/agent-project
• Author: Clayton Young / Superior Byte Works, LLC (@borealBytes)
• License: Apache-2.0

This skill (as part of research-agent-skills) is distributed under the MIT License. The included Apache-2.0 content is compatible for downstream use with attribution retained, as preserved in the file headers throughout this skill.

---

Agent operating procedure

1. Check the environment. Collect the user's draft, data, figures and target venue before writing anything.
2. Pin down the inputs. Confirm formats, identifiers and parameters from the data or the user. Ask rather than guess any value that changes the result.
3. Run a small version first. Draft one section or paragraph and confirm tone, structure and terminology with the user.
4. Execute the full task using the instructions and references above.
5. Validate the result. Every claim traces to the user's data or a verified source; numbers match tables and figures; citations are real (citation-verification).
6. Report. State what was run (versions, commands, parameters), what was checked, and what is still uncertain.

If this happens	Do this
A claim has no supporting data or citation	Flag it in the text (e.g. [citation needed]) instead of inventing support.
A function, flag or endpoint in these instructions is missing in the installed version	Check the installed version's own documentation (help(), --help, official docs), adapt, and tell the user. Never invent an API.
A required input, identifier or parameter is ambiguous	Ask the user, or state the assumption explicitly before running.

Integrity rules

• Never fabricate results, parameters, identifiers, citations or statistics. If something cannot be run or verified, say so plainly.
• The user is the author: never add results, references or claims they did not provide or approve.
• Treat version-specific details here as possibly outdated: confirm them against the official documentation for the installed version.
• Ask before actions that cost money, consume shared GPUs or cloud quota, touch personal or patient data, or cannot be undone.

Related skills

• literature-review: Runs systematic literature reviews by searching PubMed, arXiv, bioRxiv, and Semantic Scholar (plus web search via parallel-cli), screening…
• academic-plotting: Generates publication-quality figures for ML papers from research context.