Extract Clean Content from Web Pages

Defuddle is a skill that uses the Defuddle CLI to extract clean, readable content from web pages. It removes navigation, ads and other clutter, which reduces token usage. The skill's description says to use it instead of WebFetch when the user provides a URL to read or analyze, such as online documentation, articles and blog posts. It comes from the kepano/obsidian-skills repository.

The skill is a short command reference. Always use the markdown flag for markdown output. To parse a page to markdown, run defuddle parse with the URL and the markdown flag. To save the result to a file, add an output option. To extract a specific metadata property, use the property option with a name such as title, description or domain.

Use it when the user provides a normal webpage URL to read, summarize or analyze, when token efficiency matters and a noisy page-fetch approach would waste context, and for docs, articles, blog posts and similar public web content.

It is designed for standard, public web pages. It is a command-line tool, so it needs the Defuddle CLI installed. If it is not installed, the skill says to install it globally with npm. Its limitations say to use the skill only when the task clearly matches its scope, not to treat output as a substitute for validation or expert review, and to ask for clarification when inputs are missing.

The input is a page URL. Output format is chosen with a flag:

• The markdown flag gives markdown and is the default choice.
• The JSON flag gives JSON with both HTML and markdown.
• No flag gives HTML.
• The property flag with a name returns a specific metadata property.

The usage commands are:

defuddle parse <url> --md

Saving to a file uses defuddle parse with the URL, the markdown flag and the output option with a file name such as content.md. Metadata extraction runs defuddle parse with the URL and the property option, for title, description or domain.

If Defuddle is not installed, it gives one line for it, which is to run npm install -g defuddle.

It is for people and agents that read a lot of web content and want compact, clean markdown instead of full HTML pages, especially for documentation and articles, to save context and tokens.