Obsidian Clip

Workflow

1. Fetch the page
   • Prefer lightweight extraction tools (e.g., web_fetch) first.
   • Use a real browser only when needed (JS-heavy sites / WeChat / login / extraction fails).
   • If blocked by login/paywall, ask the user to log in, then retry.
2. Rewrite for readability
   • Don’t make the output about tool limitations.
   • If details are still inaccessible, produce a useful clip anyway: page定位 + 可用信息 + what you need to finish.
   • Do not send screenshots/images unless the user explicitly requests.
3. Save to Obsidian using the bundled script.

Bundled script

Path: scripts/clip_save.sh

Example:

bash scripts/clip_save.sh \
  --url "https://example.com" \
  --title "页面标题" \
  --theme "一句话主题" \
  --bullets "要点1" --bullets "要点2" \
  --actions "我怎么用1" \
  --limits "规则/限制（可选）" \
  --tags "clip" --tags "ai" \
  --keywords "keyword1-keyword2" \
  --date "YYYY-MM-DD"

Notes:

• --bullets / --actions / --limits / --tags can be repeated.
• Keep --keywords short (2–5 tokens joined by -).