web_read_page · wangzan101, tools — Claude Code prompt for reading web pages
The web_read_page Claude Code prompt extracts readable text from a web page, including content beyond the visible area. It can read the current tab or an address supplied in the request.
Apache-2.0 ★ 9,488 updated 2026-11-16
.claude/commands/web_read_page.mdWhat web_read_page does
The prompt gathers page text, headings, up to 50 links, and up to 20 images. It reads open dialogs first, then the main content. Navigation, footers, and other link-list sections are collapsed; set unfold=true to retrieve those sections in full.
Long pages can be read in parts by passing nextOffset along with the same unfold value. Without a URL, it reads the current tab. With a URL, it opens that address in a temporary background tab and leaves the current tab unchanged. The result includes a character count, the share of text that is not link text, and a verdict such as thin or image_text.
What it helps with
- extract text beyond the visible part of a page
- read open dialogs and the page’s main content
- continue reading a long page from a given offset
- identify pages with little text or image-based content
Example prompts for the agent
- Read the full page in the current tab, including any open dialogs.
- Read the page at this address and continue in parts if needed.
- Include the full navigation and footer text.
How to install web_read_page
Manually: put the files in place
.claude/commands/web_read_page.mdHow to use
Paste the text at the start of a chat or into project instructions
Questions about web_read_page
How do I install web_read_page in Claude Code?
Add the prompt file at .claude/commands/web_read_page.md.
What does web_read_page do?
It extracts readable page text, headings, links, and images, then reports how the read went. Long pages can be retrieved in parts.
What do web_read_page’s reading verdicts mean?
thin means little text was found, perhaps because the page is still loading or the content is elsewhere. needs_sign_in marks a sign-in form, refused a verification, refusal, or HTTP error, and image_text content made mostly of images.
Source text
web_read_page.md · Apache-2.0
Read the whole readable text of a web page, not limited to what is on screen: open dialogs first (between <<dialog>> and <<end of dialog>>), then the main content (headings prefixed with #), plus up to 50 links and 20 images. Navigation, footer and other link-list blocks are collapsed to one <<collapsed …>> line each; pass unfold=true to get them in full. Long pages come in parts: pass nextOffset as offset to read on, with the same unfold. Without url it reads the current tab; with url it reads that address in a temporary background tab and leaves the current tab as it is. The result tells how the read went: chars, textDensity (share of the text that is not link text) and verdict: ok; thin (little text: still loading, or the content is elsewhere); needs_sign_in (a sign-in form stands in front of the content); refused (a verification, refusal or HTTP error page); image_text (the content is mostly pictures or drawing, which text alone misses). published and updated appear when the page declares them.
© wangzan101 · Apache-2.0 · wangzan101/Mika
Similar prompts
Prompts catalog →web_read_page by other authors comes first.