Agent-Skills.md

Agent Skills: Markdown Dot New

Retrieve the Markdown version of a public web page from a URL using the agent's built-in URL fetch capability and markdown.new. Use when a user asks for page content in Markdown, cleaner extracted page content for LLM use, or URL-to-Markdown conversion.

UncategorizedID: ceshine/ceshine-agent-skills/markdown-dot-new

Install this agent skill to your local

pnpm dlx add-skill https://github.com/ceshine/ceshine-agent-skills/tree/HEAD/markdown-dot-new

Skill Files

Browse the full folder contents for markdown-dot-new.

Download Skill

Loading file tree…

markdown-dot-new/SKILL.md

Skill Metadata

Name
markdown-dot-new
Description
Retrieve the Markdown version of a public web page from a URL using the agent's built-in URL fetch capability and markdown.new. Use when a user asks for page content in Markdown, cleaner extracted page content for LLM use, or URL-to-Markdown conversion.

Markdown Dot New

Convert a public web page URL into Markdown by fetching through markdown.new.

Use This Skill When

  • The user wants a public page converted into Markdown.
  • The user wants cleaner extracted content than a normal page fetch.
  • The goal is content extraction, not interactive page inspection.

Do Not Use This Skill When

  • The page requires login or other private access.
  • The user needs to inspect live UI behavior or page structure.
  • The input URL contains query parameters. This skill only supports the GET form https://markdown.new/<target-url>, and URLs with query strings would conflict with markdown.new's own query parameters.

Workflow

  1. Validate input URL.
  • Confirm the user provided a full public URL (for example, https://example.com/page).
  • If missing or malformed, ask for a valid URL.
  • If the URL contains any query parameters, report that this skill does not support them and stop.
  1. Confirm URL-fetch tool availability.
  • Check whether the current agent/runtime can fetch URLs with its built-in fetch/browse tool.
  • If URL fetch is unavailable, tell the user that this environment cannot fetch URLs, so the conversion cannot be performed, then stop.
  1. Request Markdown from markdown.new.
  • Build a conversion URL as https://markdown.new/<target-url>.
  • Fetch that conversion URL with the built-in fetch tool.
  • Start with the default method unless the user requests a specific one.
  • Retry with method=browser only when the initial request returns an empty result or something that indicates the page is protected or requires JavaScript.
  • Retry with retain_images=true only when it becomes clear after the initial request that the page is image-rich and the images likely contain important context.
  1. Return result.
  • Provide a concise excerpt or summary by default when the Markdown is long, and offer to continue or return more.
  • Return the full Markdown only when the content is short enough or the user explicitly asks for it.
  • If markdown.new returns partial or degraded output, explain that the extraction may be incomplete.
  • If markdown.new returns an error (for example 429 or unsupported/protected page), report the error clearly and stop.

Optional Parameters

Use only when requested:

  • method=auto|ai|browser
  • retain_images=true|false

Example with parameters: https://markdown.new/https://example.com?method=browser&retain_images=true

Examples

  • "Convert this article to Markdown."
  • "Fetch this docs page as Markdown so I can quote it."
  • "Retry that extraction with browser mode because the first result was empty."

Failure Policy

Stop execution and report to the user when any of these occur:

  • No built-in URL fetch capability is available.
  • The input URL contains query parameters.
  • The URL is invalid or not publicly reachable.
  • markdown.new returns an error response.