defuddle

name	defuddle
description	Extract readable web content, removing navigation/ads to save tokens.
argument-hint	URL [--property NAME]
model	haiku

Defuddle

Instructions

1. Validate: If URL ends in .md, use WebFetch directly and return content

2. Parse arguments:

   • --property <name> → Extract metadata (title, description, domain, author, date)
   • --json → Return full JSON output
   • (default) → Return markdown content

3. Execute: npx -y defuddle parse <URL> --md (or with flags as specified)

4. Report:

   • Default: Print markdown content to stdout
   • Property mode: Print metadata value
   • JSON mode: Print full JSON structure

5. Edge cases:

   • .md URL: Skip defuddle, use WebFetch
   • Network error: "Failed to fetch content from "
   • Parse failure: "Defuddle could not extract content"
   • Missing property: "Property '' not found"

Usage Examples

Goal	Command
Parse to stdout	npx -y defuddle parse <url> --md
Get Metadata	npx -y defuddle parse <url> -p <property>
JSON Output	npx -y defuddle parse <url> --json

Options

• -p <name>: Metadata (title, description, domain, author, date)
• -o <file>: Save output to local file
• --json: Full JSON output
• --md: Markdown output (default for this skill)