Webpage Capture Markdown

Webpage Capture Markdown

Use this skill when a webpage should become a local working artifact, not just a one-off answer.

Workflow

1. Decide whether capture is necessary:
   • if the user only needs a factual answer, use normal web research
   • if the user wants a reusable file, local note, or preserved source snapshot, continue
2. Choose the lightest capture method that will work:
   • simple public page: use web tools first
   • JavaScript-heavy or interaction-heavy page: use playwright
   • authenticated or manually prepared page: ask the user before attempting interactive capture
3. Save output as markdown with minimal frontmatter:
   • url
   • title
   • captured_at
   • optional source_method
4. Default output location, unless the user specifies another path:
   • web-captures/<domain>/<slug>.md
5. Preserve the important content, headings, links, and any notes about capture limitations.
6. Do not auto-download images or media unless the user explicitly asks.
7. In the final report, include:
   • saved file path
   • capture method used
   • anything that could not be captured cleanly

Team fit

• This skill is the current Codex-workspace adaptation of a useful workflow seen in baoyu-url-to-markdown.
• Keep it toolchain-native to this workspace; do not assume bun or Claude-specific plugin behavior.