mindmap-skills

# 🧠 mindmap **Understand anything at a glance β€” turn a file, a URL, or a topic into a zoomable mindmap without leaving your AI coding agent.** [English](/mindmap-skills/) Β· [δΈ­ζ–‡](/mindmap-skills/README.zh.html) [![License: MIT](https://img.shields.io/badge/License-MIT-22c55e.svg)](LICENSE) [![Works with](https://img.shields.io/badge/works%20with-Claude%20Code%20%7C%20GitHub%20Copilot-2563eb)](#-install) [![Output](https://img.shields.io/badge/output-Markmap%20.md%20%2B%20.html-f97316)](https://markmap.js.org) [![Skills](https://img.shields.io/badge/skills-%2Fmindmap%20%7C%20%2Fmindmap--zh-8b5cf6)](#-install)

mindmap is a plugin for Claude Code and GitHub Copilot. Point it at a dense report, a long article, or just a topic, and it distills the key ideas into a clean, zoomable Markmap mindmap β€” right from your terminal.

Two languages in one plugin: /mindmap (English) and /mindmap-zh (δΈ­ζ–‡).

From any source to a map β€” in one command

flowchart LR
    IN["Input<br/>file / URL / text / topic"] --> CL{Classify}
    CL -->|URL| FE[Fetch page]
    CL -->|file| RD[Read file]
    CL -->|prose| TX[Use as text]
    CL -->|topic| KN[Model knowledge]
    FE --> ST["Build hierarchy<br/>4-7 branches, depth 3-4"]
    RD --> ST
    TX --> ST
    KN --> ST
    ST --> WR["Write Markmap .md"]
    WR --> RN{--render?}
    RN -->|yes| HT["Standalone .html"]
    RN -->|no| MD[".md deliverable"]

πŸ‘€ See it

One command turns a long GitHub blog post into a mindmap:

/mindmap https://github.blog/ai-and-ml/how-we-made-github-copilot-cli-more-selective-about-delegation/ --render

A ~1,500-word article becomes a structure you can read in seconds:

# Smarter Subagent Delegation
β”œβ”€β”€ The Problem
β”‚   β”œβ”€β”€ Delegation is powerful but not free
β”‚   └── Unnecessary handoffs, overlapping searches, waiting
β”œβ”€β”€ The Approach
β”‚   β”œβ”€β”€ Analyze β†’ find the delegation bottleneck
β”‚   β”œβ”€β”€ Change β†’ handle focused work directly
β”‚   └── Validate β†’ offline, then online, then ship
β”œβ”€β”€ Results
β”‚   β”œβ”€β”€ Tool failures per session βˆ’23%
β”‚   └── Wait time βˆ’5% P95, no quality regression
└── What's Next

That’s real output β€” see examples/ for the full Markmap .md plus the rendered interactive .html (open in any browser to zoom and collapse branches).

β–Ά Open the live interactive mindmap β€” no install, just click.


✨ Why use it

Good for: skimming research papers and reports Β· digesting blog posts and docs Β· outlining a topic before you write Β· turning meeting notes into a shareable map.


πŸ“¦ Install

The same repo is a valid plugin for both Claude Code and GitHub Copilot β€” they share the plugin/marketplace format.

Claude Code

/plugin marketplace add https://github.com/galiacheng/mindmap-skills.git
/plugin install mindmap@mindmap-marketplace
/reload-plugins

The explicit https:// URL avoids an SSH clone. The galiacheng/mindmap-skills shorthand also works if you have GitHub SSH keys configured; otherwise it fails with Permission denied (publickey).

GitHub Copilot

copilot plugin marketplace add galiacheng/mindmap-skills
copilot plugin install mindmap@mindmap-marketplace

Then run /mindmap in your session. On GitHub Copilot the skill uses the same workflow; tool names map automatically (see skills/mindmap/references/copilot-tools.md).

The marketplace manifest lives at .claude-plugin/marketplace.json.

Manual (local, no marketplace)

Copy the skill into your project’s (or user-level) skills directory:

# project-local
mkdir -p .claude/skills
cp -r skills/mindmap .claude/skills/

# or user-level (available in every project)
mkdir -p ~/.claude/skills
cp -r skills/mindmap ~/.claude/skills/

Then run /reload-skills (or restart Claude Code). Confirm with /help that /mindmap is listed.


⚑ Usage

/mindmap <input> [--panel] [--render] [--output <path>]
Input Example Behavior
File /mindmap report.md Reads the file, mirrors its structure, condenses nodes. Writes report.mindmap.md.
URL /mindmap https://example.com/article Fetches the page (WebFetch), maps its content. Writes <page-slug>.mindmap.md.
Pasted text /mindmap "notes about X, Y, Z..." Extracts concepts into branches. Writes <title-slug>.mindmap.md.
Topic /mindmap "vector databases" Generates a map from the model’s knowledge. Writes vector-databases.mindmap.md.

Flags


πŸ§‘β€βš–οΈ The judge panel (--panel)

For complex or high-stakes sources β€” papers, long reports β€” add --panel. Instead of designing the structure in a single pass, the skill runs a multi-agent panel: 3 proposers β†’ 3 judges β†’ 1 synthesizer. They compete on structure and pick the best format for every node.

flowchart TB
    SRC["Source content + skill rules"]
    subgraph PROPOSE["1 - Propose (3 lenses)"]
        P1["Proposer A<br/>narrative-first"]
        P2["Proposer B<br/>data-first"]
        P3["Proposer C<br/>audience-first"]
    end
    subgraph JUDGE["2 - Judge (score 10 dimensions)"]
        J1["Judge 1"]
        J2["Judge 2"]
        J3["Judge 3"]
    end
    subgraph SYNTH["3 - Synthesize"]
        SY["Top proposal = spine<br/>+ graft judges' best ideas"]
    end
    SRC --> P1 & P2 & P3
    P1 & P2 & P3 --> J1 & J2 & J3
    J1 & J2 & J3 --> SY
    SY --> OUT["Final Markmap .md"]

1 Β· Propose β€” three agents each design a full structure through a different lens:

Lens Designs the map around…
Narrative-first the source’s own arc and section order
Data-first the headline numbers and evidence, every metric bolded
Audience-first the punchline first, every leaf legible on a slide

Each proposer also tags every node with a format and a render tier:

Format Tier Best for
bullet list core parallel facts
bold inline core headline metric, emphasis
link core references, repo, arXiv
table rich comparisons, benchmark numbers
code block rich formulas, reward functions, code
checkbox rich tasks, limitations, checklists

2 Β· Judge β€” three judges independently score every proposal 1–5 on ten dimensions (branch count, depth, phrasing, legibility, source fidelity, punchline-first, headline numbers, leaf legibility, visual balance, format fit) and name each proposal’s single best idea.

3 Β· Synthesize β€” one agent takes the top-scored proposal as the spine, grafts in the best ideas the judges flagged from the other two, and emits the final Markmap .md.

Cost: the panel spawns ~7 agents and is token-intensive β€” reserve it for sources worth the spend. Without --panel, the skill does a fast single-pass build. The exact prompts and schemas live in skills/mindmap/references/judge-panel.md.

Both real examples in examples/ were generated with --panel.


πŸ—‚οΈ Output format

The skill writes Markmap-flavored markdown β€” a single # root, ## branches, and nested bullets:

---
title: Retrieval-Augmented Generation
markmap:
  colorFreezeLevel: 2
  maxWidth: 300
---

# Retrieval-Augmented Generation

## Indexing
- Chunk documents
- Embed chunks
- Store vectors

## Retrieval
- Embed the query
- Find nearest chunks

## Generation
- Inject context into prompt

Viewing the .md:


🌐 Rendering to HTML

/mindmap "transformer attention" --render

This runs npx markmap-cli <file>.md -o <file>.html --no-open under the hood (via skills/mindmap/scripts/render.sh).

Requirement: Node.js (provides npx). No global install needed β€” npx fetches markmap-cli on demand.


πŸ› οΈ Contributing

Curious how the skill is wired, or want to run the tests? See CONTRIBUTING.md for the project layout, the prompt-driven design, and the bash test suite.


πŸ“„ License

MIT Β© 2026 Haixia Cheng