Adobe's browser skills: get past cookie banners and bot protection, then map a page's layout, skeleton, assets and languages.
by Adobe· Web & browserUnclaimed · Yours? Claim it
4 images
About
Adobe's own plugin of web skills for coding agents. Its skills drive a headless browser through Adobe's web CLI and playwright-cli to analyse pages, clear away cookie banners and bot protection, and pull out a page's structure and assets, often as the first step of a migration to AEM Edge Delivery Services. The skills load when a request matches. What its skills cover - Probing a URL for CDN bot protection (Akamai, Cloudflare, DataDome, AWS WAF) and writing a browser recipe the other commands reuse - Detecting and dismissing cookie and consent banners, modals, newsletter popups, paywalls and login walls - Capturing a page's rendered layout as a spatial tree, with positions, sizes, grids and backgrounds - Reducing a page to a structural skeleton (skeleton.html and manifest.json) for block generation - Extracting icons, metadata and Open Graph tags, body text, forms, videos and social links - Detecting the languages a page declares and the ones its text is really in, and where they disagree - Listing every URL on a site from robots.txt and sitemaps, or by following links - Caching a site through a local proxy so it can be replayed offline - Attaching to a Chrome you already have running, and testing an unpacked Chrome extension's side panel, popup or options page - Serving a real URL behind a clean demo domain with trusted HTTPS (macOS only) How to install it In Claude Code, run /plugin marketplace add adobe/skills, then /plugin install web@adobe-skills. GitHub Copilot CLI reads the same marketplace: copilot plugin marketplace add adobe/skills, then copilot plugin install web@adobe-skills. Good to know - The skills need Node 22 or later and playwright-cli on your PATH (npm install -g @playwright/cli@latest). - Adobe's web CLI ships inside the plugin, and each command prints one JSON document. - Domain masking needs mkcert and sudo, and runs on macOS only. - The skills tell your agent to treat what it reads on web pages as untrusted content.
Once it’s added
“This page shows "access denied" in headless Chrome. Probe its bot protection.”
“Dismiss the cookie banner on example.com before you screenshot it.”
“Reduce this page to a structural skeleton for Edge Delivery Services.”
“Extract the icons, metadata and forms from this page.”
“What language is this page in, and do its hreflang links match?”
“List every URL on example.com from its sitemaps.”
Before you add it
Reads only
Web pages you point it at, in a headless browser · A site's robots.txt and sitemaps · A page's layout, languages, metadata and assets
Makes changes · your agent can ask first
Writes JSON reports, browser recipes, skeleton.html and extracted icons to your project · Dismisses cookie banners and popups in its own browser session · Adds a hosts entry and a local certificate when you mask a domain (macOS, with sudo)
From people who use it
Step by step
Nothing new reaches your agent until a person has looked at it.
No reviews yet. Used Adobe Web? Say how it went.
Write a reviewRuns on your machine.
Runs on your machine.
/plugin marketplace add adobe/skills /plugin install web@adobe-skills
Scanned and passed · checked 6 h ago
Plugins people install alongside Adobe Web.
Scanned on every change, and again when our rules improve.
Claude CodeSupported
OpenClawSupportedTell us and a person on the trust team will look at it. Or email [email protected].
Nobody has shared how they use Adobe Web yet.
Security scan
Passed
Safe to install in every agent it supports. Claude plugins.
Plugins people install alongside Adobe Web.
Hand web tasks to a Browser Use agent in a cloud browser: it browses sites, fills in forms and pulls out the data you asked for.
ByteDance's local MCP server that lets your agent drive a real browser: open pages, click, fill in forms, read pages and take screenshots.
Microsoft's MCP server for browser automation with Playwright: your agent navigates, clicks and fills in pages from structured snapshots.
Plans and builds scraping pipelines with Firecrawl or local Python, and won't hand over the data until it passes a validation check.
