Defuddle Web Extractor
Turn standard web pages into clean Markdown by removing navigation and clutter before an agent reads or analyzes them.
The skill appears limited to extracting a user-specified web page and shows no malware or credential-theft behavior. However, it requests a global npm installation and external URL access without disclosing data flow, sensitive-content handling, user confirmation, permission isolation, or rollback, so points are deducted.
Commands, parameters, and output formats are mostly internally consistent, and the .md URL exception is explicit. Dependency versions are not pinned, and there is no diagnostic or recovery guidance for installation failure, network errors, invalid URLs, dynamic pages, or parsing failures; the static cap keeps this below 10.
The audience and trigger are reasonably clear: use it for standard web-page URLs and avoid it for .md URLs. Boundaries for non-web URLs, authenticated pages, dynamic pages, and restricted pages are unspecified, as are Chinese-language and mainland-China network considerations, so points are deducted.
The documentation is readable and includes installation, core usage, file output, metadata queries, an output-format table, and MIT license context. It lacks dependency versions, skill versioning, changelog, maintenance ownership, update path, troubleshooting, and known limitations, preventing a high score.
The goal is clear and the markdown extraction command produces a directly usable target format, with a stated token-saving benefit. However, the files contain no committed tests or reproducible execution evidence, so extraction completeness, failure behavior, and comparative benefit cannot be confirmed statically; the static cap applies.
Concrete CLI commands, parameter examples, and README attribution to Defuddle provide limited auditability. There is no test suite, CI coverage, pinned dependency, or third-party execution evidence for independent reproduction, so the score remains low.
- The global installation is not version-pinned, creating version-drift and supply-chain concerns; consider pinning and documenting the dependency source.
- Users should assess whether the target page contains sensitive information and disclose what requests or content are sent to external services or dependencies.
- No fallback or clear failure guidance is provided for unreachable networks, authenticated pages, dynamic rendering, or extraction errors.
What does this skill do, and when should you use it?
This skill uses the Defuddle CLI to extract readable content from standard web pages and output Markdown. It removes navigation, ads, and other page clutter. It is intended for cases where a user provides a web URL for reading or analysis. URLs ending in `.md` should instead be handled directly with WebFetch.
Runs Defuddle CLI against a web-page URL; uses --md for Markdown output; can save the result to content.md; and can extract individual metadata fields such as title, description, or domain.
- A user needs online documentation read after navigation and advertising have been removed.
- A researcher wants a cleaner Markdown version of an article or blog post for analysis.
- A user provides a standard web-page URL and wants to reduce clutter and token usage.
- A user needs to preserve extracted page content locally as a Markdown file.
What are this skill's strengths and limitations?
- Produces Markdown that is convenient for downstream reading and analysis.
- Removes navigation, ads, and other page clutter.
- Supports direct output, file saving, and individual metadata extraction.
- Uses a plain SKILL.md structure with only `name` and `description` frontmatter fields.
- Requires npm and the Defuddle CLI to be installed.
- Scoped to standard web pages; `.md` URLs are explicitly out of scope.
- The supplied material does not document operating-system coverage, a test suite, or detailed error handling.
How do you install this skill?
Install the Defuddle CLI with npm install -g defuddle. This skill is one of five bundled in kepano/obsidian-skills; the README also documents marketplace, npx skills, Claude Code, Codex, and OpenCode installation methods for the full collection.
How do you use this skill?
For a standard web page, run defuddle parse <url> --md. Save output with defuddle parse <url> --md -o content.md. Extract metadata with defuddle parse <url> -p title. Do not use this skill for URLs ending in .md; use WebFetch directly for those URLs.
How does this skill compare with similar options?
The skill explicitly prefers Defuddle over WebFetch for standard web pages because Defuddle removes navigation, ads, and clutter; .md URLs should instead use WebFetch.