AgentSkillsCN

get-markdown

当用户提供了用于获取文档或网页内容的 URL 时，可使用此技能。系统会自动将 URL 转换为原始的 Markdown 版本，从而减少上下文窗口的占用，并有效去除 HTML 噪声。当您收到来自 GitHub、Claude/Anthropic 文档、Gemini CLI 文档、Firebase 文档、Google 开发者文档，或 OpenAI 文档的 URL 时，可使用此技能。

SKILL.md
--- frontmatter
name: get-markdown
description: This skill should be used when the user provides a URL to fetch documentation or web content. Automatically converts URLs to their raw markdown versions to reduce context window usage and eliminate HTML noise. Use when given URLs from GitHub, Claude/Anthropic docs, Gemini CLI docs, Firebase docs, Google dev docs, or OpenAI docs.

Fetch the markdown version of a URL to save context and reduce noise.

Process

  1. •Identify the URL pattern from the table below
  2. •Transform to the markdown URL
  3. •Fetch using WebFetch
  4. •Warn if content is a 404 page or cannot be read

URL Transformation Rules

SourcePatternTransformationExample
GitHubgithub.com/{owner}/{repo}/blob/{branch}/{path} (text files: .md, .txt, .json, .adoc, .rst, etc.)raw.githubusercontent.com/{owner}/{repo}/refs/heads/{branch}/{path}github.com/google-gemini/gemini-cli/blob/main/docs/cli/headless.md → raw.githubusercontent.com/google-gemini/gemini-cli/refs/heads/main/docs/cli/headless.md
Claude Code docscode.claude.com/docs/en/{page} or code.claude.com/docs/en/{page}#{anchor}code.claude.com/docs/en/{page}.md (strip anchor)code.claude.com/docs/en/skills#pass-arguments-to-skills → code.claude.com/docs/en/skills.md
Anthropic API docsplatform.claude.com/docs/en/{path}platform.claude.com/docs/en/{path}.mdplatform.claude.com/docs/en/api/overview → platform.claude.com/docs/en/api/overview.md
Gemini CLI docsgeminicli.com/docs/{path}First convert to github.com/google-gemini/gemini-cli/blob/main/docs/{path}.md, then apply GitHub rulegeminicli.com/docs/cli/headless/ → raw.githubusercontent.com/google-gemini/gemini-cli/refs/heads/main/docs/cli/headless.md
Firebase docsfirebase.google.com/docs/{path}firebase.google.com/docs/{path}.md.txtfirebase.google.com/docs/ai-logic → firebase.google.com/docs/ai-logic.md.txt
Google dev docsai.google.dev/{path}ai.google.dev/{path}.md.txtai.google.dev/gemini-api/docs → ai.google.dev/gemini-api/docs.md.txt
OpenAI docsplatform.openai.com/docs/{path}platform.openai.com/docs/{path}.mdplatform.openai.com/docs/guides/prompt-engineering → platform.openai.com/docs/guides/prompt-engineering.md

Edge Cases

  • •Trailing slashes: Strip before transformation (e.g., geminicli.com/docs/cli/headless/ → headless)
  • •URL anchors: Strip #anchor fragments before adding .md extension
  • •Query parameters: Strip ?params before transformation
  • •Unknown patterns: If URL does not match any pattern, fetch as-is with WebFetch and inform user that no markdown conversion was available

Error Handling

After fetching, check for:

  1. •404 responses: Warn "The markdown URL returned a 404. The page may not exist or the URL pattern may have changed."
  2. •Empty content: Warn "The markdown URL returned empty content."
  3. •HTML instead of markdown: If response contains <!DOCTYPE or <html, warn "The URL returned HTML instead of markdown. The transformation pattern may be incorrect."

Output

After successful fetch, present the content to the user. If there were warnings, display them prominently before the content.