Best AI Skills for Browser Automation in 2026: 12 Compared
Top 12 browser-automation skills for Claude Code, ranked. Five are Playwright wrappers; the differentiated picks are accessibility-tree, native-UI, and API-backed.
Browser automation is the most fragmented category in the catalog. Twelve top skills, 329.8K combined signal — and five of them are Playwright wrappers with minor feature differences. The real story is the non-Playwright differentiation: accessibility-tree-based agents that don't hallucinate selectors, native-UI control for macOS, anti-detection stacks for sites that actively fight bots, and API-backed alternatives that skip the browser entirely. We ranked the top 12 because picking the right one is more about which mechanism matches your problem than picking the most-installed entry.
Quick Pick
Agent Browser — A11y-tree-based headless browser CLI built specifically for AI agents. The accessibility-tree snapshot + ref-based selectors prevents the "Claude invented a CSS selector that doesn't exist" failure mode that plagues vanilla Playwright wrappers. 66.5K signal, top of the category for a reason.
What These Skills Actually Do
Browser automation skills fall into five mechanisms, and picking the right mechanism matters more than picking the most popular entry: (1) Accessibility-tree (Agent Browser #1) — snapshots semantic structure, AI selects by ref, far fewer selector hallucinations. (2) Native-UI (Desktop Control #2, Peekaboo #7) — mouse/keyboard/screen at the OS level, used for non-browser apps or when you specifically need the visual rendering. (3) Playwright wrappers (#3, #4, #5, #6, #8, #10) — DOM-level automation; great when the site has clean structure, error-prone when it doesn't. (4) API-backed (Firecrawl #9) — no local browser at all; you pay a vendor to handle the scraping and just get structured data back. (5) Anti-detection (Stealth Browser #12, Fastest Browser Use #11) — for sites that aggressively fingerprint headless browsers. The mediocre skills are the Playwright wrappers that don't add anything; the great ones either pick a different mechanism or stack meaningful extras (stealth, multi-tab orchestration, persistent sessions) on top.
How We Ranked
We sorted 12 candidate skills by a composite score:
- Popularity signal — the highest of GitHub stars, install count, or ClawHub download count. Log-scaled so a 100-star skill doesn't get buried under a 100,000-star one if the smaller one is meaningfully better.
- Quality score — when set, a 0–5 rubric that breaks ties within popularity tiers. Roughly 15% of catalog skills carry a quality score today; we surface it in the comparison table when available.
The formula is identical across the entire Best-Of 2026 series, so you can compare apples to apples between categories.
The Top 12
1. Agent Browser
Skill · matrixy/agent-browser-clawdbot · 66.5K signal · quality unrated Headless browser automation CLI optimized for AI agents with accessibility tree snapshots and ref-based element selection.
The take: The most important architectural choice in the category. Instead of "find the button with class .btn-primary," it gives Claude an a11y tree and a ref ID. Claude can't hallucinate selectors that don't exist because there's nothing to hallucinate — only refs that were in the snapshot. Install this if you've ever watched Claude invent a CSS selector and waste 10 minutes debugging.
2. Desktop Control
Skill · matagul/desktop-control · 42.8K signal · quality unrated Advanced desktop automation with mouse, keyboard, and screen control.
The take: Not browser automation — OS-level automation. Use this when the target isn't a website (native macOS/Windows app, legacy enterprise software with no API). The 42.8K signal means a lot of people have apps that don't have headless equivalents. Slower and more brittle than browser automation by design, but sometimes the only option.
3. Browser Use
Skill · shawnpana/browser-use · 33.3K signal · quality unrated Automates browser interactions for web testing, form filling, screenshots, and data extraction.
The take: First of the Playwright wrappers and the highest-installed. Generic feature set — navigation, forms, screenshots, extraction. Beats the next four wrappers (#4, #5, #6, #10) on install signal but not on any specific differentiator. The fact that the most popular Playwright wrapper isn't materially better than the others is the data point.
4. Browser Automation
Skill · peytoncasper/browser-automation · 31.3K signal · quality unrated Natural-language CLI for browser actions — browse, navigate, extract, screenshot, fill, click.
The take: Functionally near-identical to #3. Picks "natural-language CLI" as its framing, which is mostly a marketing tagline at this point — every wrapper accepts NL prompts from Claude. Choose between #3 and #4 by reading the README and seeing which API feels cleaner; you won't go wrong either way.
5. Playwright MCP
Skill · spiceman161/playwright-mcp · 30.7K signal · quality unrated Browser automation via Playwright MCP server.
The take: MCP-flavored wrapper instead of CLI-flavored. The MCP framing matters if you're integrating with a multi-skill agent setup where tool discovery and orchestration matter — MCP gives Claude richer metadata about each browser primitive. For one-shot scripts, the CLI wrappers (#3, #4) are simpler.
6. Playwright (Automation + MCP + Scraper)
Skill · ivangdavila/playwright · 23.3K signal · quality unrated Playwright via MCP for navigation, click, form, screenshot, extract, debug.
The take: Same ivangdavila author behind Market Research (#6 in marketing list) and SEO — pattern of building consistent, structured skills. This entry combines automation, MCP, and scraping modes in one skill. Pick this over #5 if you also want a scraping mode without installing two skills.
7. Peekaboo
Skill · steipete/peekaboo · 22.4K signal · quality unrated Capture and automate macOS UI with the Peekaboo CLI.
The take: macOS-specific UI automation — the native-app equivalent of #2 but Mac-only and tighter on screen capture / accessibility-API integration. Use this over Desktop Control if your target is a Mac app and you care about reliability over cross-platform support.
8. Playwright Scraper Skill
Skill · waisimon/playwright-scraper-skill · 22.2K signal · quality unrated Playwright-based web scraping with anti-bot protection. Tested on complex sites like Discuss.com.hk.
The take: Scraping-optimized Playwright wrapper — adds anti-bot handling on top of the generic wrappers. The Discuss.com.hk reference is meaningful: that's a non-trivial Cantonese forum with real bot detection. If your scraping target is structurally hostile, this is the wrapper to install before reaching for #12.
9. Firecrawl Search
Skill · ashwingupy/firecrawl-search · 16.2K signal · quality unrated Web search and scraping via Firecrawl API. Search, scrape (including JS-heavy pages), crawl entire sites, extract structured data.
The take: Different mechanism — no local browser at all. Firecrawl is a hosted service that handles the browser-running for you and returns structured data. Trade-offs are obvious: you pay them, you trust them with your traffic, but you skip the whole "is the headless browser installed correctly" problem. Best fit for cron-job pipelines, worst fit for one-off interactive sessions.
10. Playwright Commander
Skill · austindixson/playwright-commander · 15.8K signal · quality unrated Playwright wrapper focused on advanced UI automation, analysis, and debugging.
The take: Yet another Playwright wrapper. Its differentiator on the README is "debugging" framing — slightly more verbose state reporting back to Claude when actions fail. Marginal advantage over #3–#6; install only if you keep hitting hard-to-debug action failures.
11. Fastest Browser Use
Skill · rknoche6/fast-browser-use · 12.7K signal · quality unrated High-performance browser automation for heavy scraping, multi-tab management, and precise DOM extraction.
The take: Performance-focused — multi-tab orchestration and persistent state (cookies/localStorage) are the actual differentiators here, not raw speed. Install when "I need to keep 30 sessions open and reuse them across runs" describes your problem.
12. Stealth Browser
Skill · mayuqi-crypto/stealth-browser · 12.6K signal · quality unrated Stealth browser automation with anti-detection, Cloudflare bypass, CAPTCHA solving, persistent sessions, and silent operation.
The take: Specialized for sites that actively fight automation. The legitimate uses are real (your own QA tests on Cloudflare-protected staging environments, monitoring competitor pricing on your own surfaces, accessibility-testing scripts). The illegitimate uses are also obvious — read the target site's ToS before pointing this at anything you don't own. Heaviest stack in the category by far.
Comparison Table
| # | Skill | Type | Stars / Installs | Quality | License |
|---|---|---|---|---|---|
| 1 | Agent Browser | Skill | 66.5K | — | — |
| 2 | Desktop Control | Skill | 42.8K | — | — |
| 3 | Browser Use | Skill | 33.3K | — | — |
| 4 | Browser Automation | Skill | 31.3K | — | — |
| 5 | Playwright MCP | Skill | 30.7K | — | — |
| 6 | Playwright (Automation + MCP + Scraper) | Skill | 23.3K | — | — |
| 7 | Peekaboo | Skill | 22.4K | — | — |
| 8 | Playwright Scraper Skill | Skill | 22.2K | — | — |
| 9 | Firecrawl Search | Skill | 16.2K | — | — |
| 10 | Playwright Commander | Skill | 15.8K | — | — |
| 11 | Fastest Browser Use | Skill | 12.7K | — | — |
| 12 | Stealth Browser | Skill | 12.6K | — | — |
FAQ
How is this list different from the category page on aiskill.market?
The category page is a directory: every skill in the category, sortable and filterable. This list is editorial — opinionated, time-stamped (2026-05-18), and ranked. Use the directory when you know what you want; use this when you don't.
Why does the #1 pick have fewer stars than #5?
Stars are one signal among several. The composite score above also includes install counts (which reflect actual usage on aiskill.market) and the optional quality score. A skill with a smaller star count can rank higher if its installs or quality score are strong enough to offset.
Are these all free?
The skills are open source. Three carry paid-service dependencies worth flagging: Firecrawl Search (#9) requires a Firecrawl API key (free tier exists, paid for volume), Stealth Browser (#12) often pairs with a CAPTCHA-solving service (2Captcha, etc.) which is paid per solve, and any cloud-rendering Playwright workflow (Lambda, browserless.io) incurs cloud bills regardless of the skill being free.
How do I install one?
Each linked skill page has install instructions. The fastest path is the one-line install via the aiskill.market CLI or by adding the source repo as a Claude Code plugin marketplace. Note: several Playwright-based skills (#3–#8, #10, #11) will install Chromium on first run — that's ~500MB of disk.
How often does this list update?
Quarterly. Browser-automation tooling churns less than creative tooling — the underlying engines (Chromium, Playwright, the OS accessibility APIs) are stable. Expect ranking shifts within the Playwright wrappers but not the fundamental mechanism breakdown.
Should I use the accessibility-tree skill or a Playwright wrapper?
Default to Agent Browser (#1) if you're building an AI workflow. The a11y-tree approach is meaningfully more reliable for AI-driven automation because it eliminates selector hallucination. Use a Playwright wrapper when you already have an existing test suite or need DOM-level control (XPath selectors, shadow DOM, etc.) that the a11y tree doesn't expose.
What about anti-bot evasion — is that ethical?
Anti-detection skills (#12, partially #8) are legitimately dual-use. Defensive use cases (testing your own sites' bot protection, monitoring scraping resilience, automating QA against your own auth flows) are fine. Adversarial use (scraping sites that explicitly prohibit it in their ToS, defeating rate-limiting on services you don't own) puts you at legal risk. The skill itself is neutral — your target matters.
Related Categories
- Best AI Skills for Development & Code Tools in 2026 — testing and CI overlap with browser automation
- Best AI Skills for Productivity & Organization in 2026 — desktop control (#2) and Peekaboo (#7) are useful here
- Best AI Skills for Business & Marketing in 2026 — scraping + social-media skills are common pairings
Browse The Full Catalog
Find every skill in this category — including the ones that didn't make the top 12 — at the Browser Automation page.
Part of the Best-Of 2026 series. Updated 2026-05-18. Skills sampled from a catalog of ~262 active entries with a combined 329.8K popularity signal across the ranked entries.