Markdown Helper Skill

Ultra-efficient markdown operations using native CLI tools. Save 60-70% tokens on common markdown tasks.

Overview

This skill provides markdown operations WITHOUT reading entire files into context, saving 68% tokens per operation.

Cross-Platform: Works on Windows, Mac, and Linux (uses Node.js + CLI tools).

Token Savings:

• Traditional approach: ~800 tokens
• This skill: ~250 tokens
• Savings: 550 tokens per query (68%)

🔧 BASH COMMAND ATTRIBUTION PATTERN

CRITICAL: Before executing EACH bash/node command, MUST output:

🔧 [markdown-helper] Running: <command>

Examples:

🔧 [markdown-helper] Running: node ~/.claude/skills/markdown-helper/md-helper.js extract-headers README.md
🔧 [markdown-helper] Running: node ~/.claude/skills/markdown-helper/md-helper.js generate-mermaid workflow.md
🔧 [markdown-helper] Running: node ~/.claude/skills/markdown-helper/md-helper.js lint *.md

Why: This pattern helps users identify which skill is executing which command, improving transparency and debugging.

🎨 VISUAL OUTPUT FORMATTING

CRITICAL: All markdown-helper output MUST use the colored-output formatter skill!

Use Colored-Output Skill

Every response MUST start with:

bash .claude/skills/colored-output/color.sh skill-header "markdown-helper" "Message here..."

IMPORTANT: Use MINIMAL colored output (2-3 calls max) to prevent screen flickering!

Example formatted output (MINIMAL PATTERN):

# START: Header only
bash .claude/skills/colored-output/color.sh skill-header "markdown-helper" "Parsing markdown file..."

# MIDDLE: Regular text (no colored calls)
Extracting headers from file...
Found 12 headings (4 H1, 6 H2, 2 H3)
Processing tables and code blocks...

# END: Result only
bash .claude/skills/colored-output/color.sh success "" "Markdown parsed successfully"

WHY: Each bash call creates a task in Claude CLI, causing screen flickering. Keep it minimal!

Usage Examples

Extract Headers

• "Extract all headers from TASK-024.md"
• "Show me the structure of README.md"
• "List all H2 and H3 headings in this file"

Extract Tables

• "Parse the tables in CHANGELOG.md"
• "Show me table data from specifications.md"
• "Extract tables as JSON from this markdown"

Generate Diagrams

• "Create a flow diagram showing the user registration process"
• "Generate a Mermaid diagram for this workflow"
• "Make a flowchart for the checkout process"

Lint & Format

• "Fix markdown formatting in project-tasks/*.md"
• "Lint all markdown files in this directory"
• "Auto-fix markdown issues in README.md"

Bulk Operations

• "Replace 'SubHero' with 'SubsHero' in all markdown files"
• "Find all TODOs in project-tasks/"
• "Update all links from HTTP to HTTPS"

Commands

1. Extract Headers

Command:

node ~/.claude/skills/markdown-helper/md-helper.js extract-headers <file>

Options:

• --level <1-6>

  - Filter by heading level (e.g.,

  --level 2

  for H2 only)

• --json

  - Output as JSON

• --count

  - Show count per level

Examples:

# All headers
node md-helper.js extract-headers README.md

# H2 headers only
node md-helper.js extract-headers README.md --level 2

# JSON output
node md-helper.js extract-headers TASK-024.md --json

Output:

📄 Headers in README.md:
═══════════════════════════════════════
H1: Markdown Helper Skill
H2: Overview
H2: Usage Examples
H3: Extract Headers
H3: Extract Tables
H2: Commands

2. Extract Tables

Command:

node ~/.claude/skills/markdown-helper/md-helper.js extract-tables <file>

Options:

• --json

  - Output as JSON array

• --index <n>

  - Extract specific table by index (0-based)

• --format <csv|json|md>

  - Output format

Examples:

# Extract all tables
node md-helper.js extract-tables CHANGELOG.md

# First table only
node md-helper.js extract-tables specs.md --index 0

# As CSV
node md-helper.js extract-tables data.md --format csv

Output:

[
  {
    "headers": ["Command", "Description", "Token Savings"],
    "rows": [
      ["extract-headers", "Parse headers", "500 tokens"],
      ["extract-tables", "Parse tables", "600 tokens"]
    ]
  }
]

3. Generate Diagram

Command:

node ~/.claude/skills/markdown-helper/md-helper.js generate-diagram <type> <output>

Types:

• flowchart

  - Flow diagrams

• sequence

  - Sequence diagrams

• class

  - Class diagrams

• gantt

  - Gantt charts

• er

  - Entity relationship diagrams

Options:

• --input <file>

  - Read Mermaid syntax from file

• --format <svg|png|pdf>

  - Output format (default: svg)

• --theme <default|dark|forest|neutral>

  - Diagram theme

Examples:

Interactive (Claude generates Mermaid syntax):

# User: "Create a flowchart for user registration"
# Claude generates Mermaid syntax, then:
node md-helper.js generate-diagram flowchart user-registration.svg

From File:

# diagram.mmd contains Mermaid syntax
node md-helper.js generate-diagram flowchart output.svg --input diagram.mmd

Mermaid Syntax Example:

flowchart TD
    A[Start] --> B{Is user logged in?}
    B -->|Yes| C[Show Dashboard]
    B -->|No| D[Show Login]
    D --> E[Authenticate]
    E -->|Success| C
    E -->|Failure| D

4. Lint & Format

Command:

node ~/.claude/skills/markdown-helper/md-helper.js lint <file-or-pattern>

Options:

• --fix

  - Auto-fix issues (default: true)

• --check

  - Check only, don't fix

• --config <file>

  - Custom config file

Examples:

# Fix single file
node md-helper.js lint README.md

# Check without fixing
node md-helper.js lint README.md --check

# Fix all markdown in directory
node md-helper.js lint "project-tasks/**/*.md"

Fixed Issues:

• Missing blank lines around headings
• Inconsistent list markers
• Trailing whitespace
• Multiple consecutive blank lines
• Heading increment violations
• And 50+ more rules

Output:

✅ Fixed 12 issues in README.md:
   • Added blank line before heading (line 45)
   • Fixed list marker spacing (line 67)
   • Removed trailing whitespace (5 lines)

⚠️  1 warning:
   • Line length exceeds 80 characters (line 123)

5. Bulk Search & Replace

Command:

node ~/.claude/skills/markdown-helper/md-helper.js replace <pattern> <replacement> <files>

Options:

• --regex

  - Use regular expressions

• --case-sensitive

  - Case-sensitive matching

• --dry-run

  - Preview changes without applying

• --backup

  - Create .bak files

Examples:

# Simple replacement
node md-helper.js replace "SubHero" "SubsHero" "**/*.md"

# Regex replacement
node md-helper.js replace "http://" "https://" "*.md" --regex

# Dry run (preview)
node md-helper.js replace "TODO" "DONE" "tasks/*.md" --dry-run

Output:

🔍 Scanning: 47 markdown files
📝 Found 23 matches in 8 files
✅ Replaced: 23 occurrences

Files modified:
   • README.md (3 replacements)
   • TASK-024.md (8 replacements)
   • CHANGELOG.md (12 replacements)

6. Extract Lists

Command:

node ~/.claude/skills/markdown-helper/md-helper.js extract-lists <file>

Options:

• --type <ordered|unordered|task>

  - Filter by list type

• --json

  - Output as JSON

Examples:

# All lists
node md-helper.js extract-lists TODO.md

# Task lists only (- [ ] items)
node md-helper.js extract-lists TODO.md --type task

# JSON output
node md-helper.js extract-lists TASKS.md --type task --json

Output:

{
  "taskLists": [
    {"checked": false, "text": "Install CLI tools"},
    {"checked": true, "text": "Create skill directory"},
    {"checked": false, "text": "Test operations"}
  ]
}

7. Statistics

Command:

node ~/.claude/skills/markdown-helper/md-helper.js stats <file>

Output:

📊 Statistics for TASK-024.md:
═══════════════════════════════════════
Lines:           487
Words:           3,245
Characters:      21,890
Headings:        23 (H1: 1, H2: 8, H3: 14)
Tables:          4
Code Blocks:     12
Links:           67
Images:          3
Lists:           15
Task Lists:      8 (6 incomplete)
Blockquotes:     2

Auto-Activation Triggers

Claude should automatically use this skill when detecting these keywords/patterns:

Primary Triggers (Auto-activate skill)

• "extract headers" / "parse headers" / "show headers"
• "extract tables" / "parse tables" / "show tables"
• "extract lists" / "parse lists" / "show task lists"
• "markdown stats" / "markdown statistics" / "file stats"
• "lint markdown" / "fix markdown formatting"
• "generate diagram" / "create flowchart" / "mermaid diagram"
• "bulk replace in markdown" / "search replace markdown files"

File Size Threshold

• Recommended: Files >500 characters (61%+ token savings)
• Optimal: Files >5 KB (84%+ token savings)
• Maximum benefit: Files >10 KB (90%+ token savings)

When NOT to Auto-Activate

• ❌ Files <100 characters (minimal savings)
• ❌ User needs content understanding, not just structure
• ❌ User explicitly says "read" or "show me the content"
• ❌ One-time operation where setup overhead isn't worth it

Operation Type Detection

✅ Use Skill For (Structure extraction):

• Headers, tables, lists, code blocks
• Statistics, metrics, counts
• Formatting, linting, validation
• Bulk operations across multiple files

❌ Use Native Read For (Content understanding):

• Reading prose/documentation to understand context
• Code review requiring full file comprehension
• Debugging specific content issues
• When file is already in context

Natural Language Processing

When User Says... Execute This:

extract-headers <file>

extract-tables <file>

generate-diagram flowchart

lint <file> --fix

replace <old> <new> <files>

stats <file>

extract-lists <file> --type task

Workflow Examples

Example 1: Document Analysis

User: "Analyze the structure of TASK-024.md"

Execute:

node md-helper.js extract-headers TASK-024.md
node md-helper.js stats TASK-024.md

Output:

Headers: 23 (H1: 1, H2: 8, H3: 14)
Key Sections:
- Business Requirements
- Technical Implementation
- Test Cases
- Acceptance Criteria

Statistics: 487 lines, 3,245 words, 67 links

Example 2: Create Flow Diagram

User: "Create a flowchart showing the checkout process with payment, validation, and order confirmation"

Claude: First, let me generate the Mermaid syntax:

flowchart TD
    A[Shopping Cart] --> B{Items in cart?}
    B -->|No| C[Show empty cart]
    B -->|Yes| D[Proceed to checkout]
    D --> E[Enter shipping info]
    E --> F[Select payment method]
    F --> G{Validate payment}
    G -->|Invalid| F
    G -->|Valid| H[Process payment]
    H --> I{Payment success?}
    I -->|No| J[Show error]
    I -->|Yes| K[Create order]
    K --> L[Send confirmation email]
    L --> M[Show order confirmation]

Execute:

echo "flowchart TD..." > /tmp/checkout.mmd
node md-helper.js generate-diagram flowchart checkout-process.svg --input /tmp/checkout.mmd

Output: SVG file created at checkout-process.svg

Example 3: Bulk Documentation Update

User: "Fix all markdown formatting issues in project-tasks/"

Execute:

node md-helper.js lint "project-tasks/**/*.md" --fix

Output:

✅ Processed 47 files
✅ Fixed 234 issues across 28 files
⚠️  3 warnings (line length)
🎉 Done! All markdown files formatted correctly.

Example 4: Extract All TODOs

User: "Find all TODO items in markdown files"

Execute:

node md-helper.js replace "TODO" "TODO" "**/*.md" --dry-run
# OR
node md-helper.js extract-lists "**/*.md" --type task --json

Output:

{
  "files": [
    {
      "file": "TASK-024.md",
      "todos": [
        {"line": 45, "checked": false, "text": "Implement coupon validation"},
        {"line": 67, "checked": true, "text": "Write unit tests"}
      ]
    }
  ]
}

Technical Details

CLI Tools Used

1. Marked CLI - Fast markdown parser

   • Converts MD to AST (Abstract Syntax Tree)
   • Supports CommonMark + GFM extensions
   • 10x faster than reading into Claude context

2. Mermaid CLI - Diagram generation

   • Flowcharts, sequence diagrams, class diagrams
   • Outputs SVG, PNG, PDF
   • 500+ diagram types supported

3. markdownlint-cli2 - Linting & formatting

   • 50+ markdown rules
   • Auto-fix most issues
   • Configurable via .markdownlint.json

Token Efficiency

Traditional Approach:

User: "Extract headers from 500-line markdown"
Claude reads entire file (500 lines × ~1.5 tokens/line) = 750 tokens
Claude parses in context = +50 tokens
Total: ~800 tokens

With Markdown Helper:

User: "Extract headers from 500-line markdown"
Claude executes: node md-helper.js extract-headers file.md
Tool output: 20 lines of headers = 30 tokens
Command overhead = 220 tokens
Total: ~250 tokens
Savings: 68%

10 operations per day = 5,500 tokens saved Monthly savings = ~165,000 tokens

Error Handling

File Not Found

❌ Error: File not found 'unknown.md'
   Check file path and try again

Invalid Mermaid Syntax

❌ Error: Invalid Mermaid syntax
   Line 5: Expected node definition

Tip: Validate syntax at https://mermaid.live/

No Tables Found

⚠️  No tables found in README.md
   This file contains no markdown tables

Configuration

Create

~/.claude/skills/markdown-helper/.markdownlint.json

{
  "default": true,
  "MD013": false,
  "MD033": false,
  "line-length": false
}

Performance

Average savings: 68-70% per operation

Maintenance

Updating CLI Tools

npm update -g marked-cli @mermaid-js/mermaid-cli markdownlint-cli2

Testing

cd ~/.claude/skills/markdown-helper
node md-helper.js --test

Migration Guide

From Manual Parsing

Before: Claude reads entire file → parses → extracts After:

node md-helper.js extract-headers file.md

From MCP Servers

Before: Use markdown MCP server (~400 tokens) After: Use this skill (~250 tokens) Savings: 37% additional tokens

Support

File Location:

~/.claude/skills/markdown-helper/

Files:

• skill.md

  - This documentation

• installation.md

  - Setup guide

• md-helper.js

  - Main Node.js script

Common Issues:

• If command not found: Check npm global path
• If module error: Run

  npm install

  in skill directory
• If diagram fails: Update Mermaid CLI

Version History

v1.0.0 (2025-10-20)

• Initial release
• Extract headers, tables, lists
• Generate Mermaid diagrams (flowcharts, sequence, class, gantt, ER)
• Lint and auto-fix markdown formatting
• Bulk search/replace operations
• Statistics and analysis
• 68% token savings vs. traditional approach
• Cross-platform support (Windows/Mac/Linux)