Changelog

0.1.2

• The extract skills no longer describe Extract as a beta or as gated by domain.

0.1.1

• Review cases checked against the live MCP server on 2026-10-13. Case 1 read scrapingcourse.com/javascript-rendering/, which scrape returns as an "Enable JavaScript" placeholder with default settings; it now reads /table-parsing, which returns its full table. Case 3 expected page titles, which batch_results doesn't return (it returns a status and a download link per URL); it now expects per-URL status.
• The starter prompt that used /javascript-rendering/ uses /table-parsing instead.

0.1.0

• First version: the Zenrows MCP server (OAuth sign-in) and nine skills: using-zenrows, getting-started, scrape-webpage, extract-structured-data, batch, crawl, map, browser-automation and capture-api-calls.
• Skills come from ZenRows/claude-zenrows 0.1.2, with Claude-specific wording replaced by provider-neutral wording, as OpenAI asks for imported Claude plugins.
• crawl uses the crawl_* tools added in zenrows-mcp 2.6.0 instead of a scrape loop. using-zenrows lists the crawl_* and account_usage tools.
• Every skill states that the user's instructions take precedence over it, as OpenAI's instruction-following guidance recommends.
• plugin.json carries the listing, five positive and three negative review cases, commerce declaration and release notes.
• scripts/validate.mjs checks the package against OpenAI's documented submission limits; scripts/package.sh builds the upload ZIP.