菜单
Browser automation for ChatGPT Plus and Gemini Advanced web interfaces. Uses Playwright MCP to interact with your paid subscriptions without API costs. Supports both models for writing comparison.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Transforms research and knowledge base material into information-dense infographic copy—clear, scannable, authoritative, and reference-worthy. Use when converting clinical algorithms, decision trees, comparison data (drug vs drug, treatment options), trial results, myth vs truth content, patient preparation guides, checklists, or sequential process workflows into structured copy ready for single-image infographic formats. Supports multiple infographic types including hero stats, dense multi-section layouts, side-by-side comparisons, myth-busters, step-by-step processes, and patient checklists. Outputs structured copy with compressed noun-heavy phrasing, specific numbers, trial citations, and JSON-formatted handoff data for downstream infographic generation.
Manages all other skills by detecting capability gaps, proposing new skills, and keeping the system registry in sync. Use when the user encounters something the system cannot do, wants to log a capability gap, review the skill backlog, propose or build a new skill, analyze recurring unmet needs, or run sync pipelines to update context files (CLAUDE.md, GEMINI.md, AGENTS.md) after skill changes.
Template-driven medical infographics using the AntV Infographic framework. Use when creating structured infographics from clinical data, trial results, or patient education content using pre-built templates for hero stats, comparison layouts, process flows, myth-busters, and checklists. Outputs publication-ready HTML/PNG infographics with JACC/NEJM-grade design tokens.
Declarative, grammar-based charting using AntV G2 for complex medical visualizations. Use when creating forest plots, Kaplan-Meier curves, funnel plots, waterfall charts, or any statistical/clinical chart that requires grammar-of-graphics composition with medical-journal-grade styling and accessibility-validated color palettes.
Publication-grade design tokens and utilities for Nature/JACC/NEJM quality graphics. Use when generating charts, infographics, animations, or any visual content that requires medical-journal-grade color palettes, typography, accessibility-validated contrast ratios, and consistent branding. Provides Python token APIs, colorblind-safe palettes, G2 chart templates, AntV infographic templates, Vizzu animation presets, and Manim integration for animated medical explainers.
Content OS orchestrator - the master skill that produces ALL content types from one seed idea (forward mode) or splits long-form content into short-form pieces (backward mode). Invokes research, writing, quality review, and visual generation skills in a coordinated pipeline. Long-form content goes through full quality gates; short-form gets quick accuracy pass.
| name | browser-automation |
| description | Browser automation for ChatGPT Plus and Gemini Advanced web interfaces. Uses Playwright MCP to interact with your paid subscriptions without API costs. Supports both models for writing comparison. |
Use your ChatGPT Plus and Gemini Advanced subscriptions through browser automation. No API costs - just your monthly subscription.
┌─────────────────────────────────────────────────────────────────┐
│ BROWSER AUTOMATION FLOW │
├─────────────────────────────────────────────────────────────────┤
│ │
│ Your Prompt ──► Playwright MCP ──► Browser Instance │
│ │ │
│ ┌────────────┴────────────┐ │
│ ▼ ▼ │
│ ┌─────────────┐ ┌─────────────┐ │
│ │ ChatGPT │ │ Gemini │ │
│ │ chat.openai│ │ gemini. │ │
│ │ .com │ │ google.com│ │
│ └──────┬──────┘ └──────┬──────┘ │
│ │ │ │
│ ▼ ▼ │
│ Response captured & returned to you │
│ │
└─────────────────────────────────────────────────────────────────┘
You have Playwright MCP configured. Verify it's working:
Use browser_snapshot to check if browser is available
The browser automation uses saved sessions. You need to log in once:
First-time setup:
Step 1: Navigate to ChatGPT
browser_navigate → https://chat.openai.com
Step 2: Check if logged in
browser_snapshot → Look for chat input or login button
Step 3: If not logged in, authenticate
browser_click → "Log in" button
browser_type → Enter email
browser_click → Continue
browser_type → Enter password
browser_click → Log in
Step 4: Start new chat
browser_click → "New chat" button (or navigate to chat.openai.com)
Step 5: Type your prompt
browser_type → Your prompt text in the message input
Step 6: Submit and wait
browser_click → Send button
browser_wait_for → Wait for response to complete
Step 7: Capture response
browser_snapshot → Get the response text
I will now use browser automation to get ChatGPT's response:
1. browser_navigate to https://chat.openai.com
2. browser_snapshot to see current state
3. browser_type to enter prompt in textarea
4. browser_click to send
5. browser_wait_for response
6. browser_snapshot to capture output
Step 1: Navigate to Gemini
browser_navigate → https://gemini.google.com
Step 2: Check if logged in
browser_snapshot → Look for chat input
Step 3: Type your prompt
browser_type → Your prompt in the input area
Step 4: Submit
browser_press_key → Enter (or click send button)
Step 5: Wait and capture
browser_wait_for → Response generation
browser_snapshot → Get response
When you want me to use browser automation, say:
"Use browser automation to ask ChatGPT: [your prompt]"
"Get Gemini's take on: [your prompt]"
"Compare browser outputs for: [your prompt]"
I will then:
Browser automation works best with persistent sessions:
# The Playwright MCP maintains browser state
# Once logged in, sessions typically persist
If you see a login screen:
If 2FA is required:
| Limitation | Workaround |
|---|---|
| Slower than API | Use for comparison, not bulk |
| Can break if UI changes | Report issues, I'll adapt |
| Requires active session | Keep browser open |
| Rate limits still apply | Don't spam requests |
| CAPTCHAs possible | May need manual intervention |
Step 1: Write with Claude (default, in this conversation)
Step 2: browser_navigate to ChatGPT, get response
Step 3: browser_navigate to Gemini, get response
Step 4: Compare all three side-by-side
"Compare how you, ChatGPT, and Gemini would write a tweet about
the cardiovascular benefits of SGLT2 inhibitors"
I will:
browser_close → Close current browser
Then start fresh with browser_navigate
browser_snapshot → Check current state
browser_navigate → Go to correct URL
browser_snapshot → Get fresh page state
Look for correct element reference
Retry with updated reference
browser_navigate → Go to login page
Complete login flow
Resume automation
This skill works with multi-model-writer:
API Models:
- /write-glm → Z.AI API
- /write-gpt → OpenAI API
- /write-gemini → Google AI Studio API
Browser Models:
- /browser-chatgpt → ChatGPT Plus web
- /browser-gemini → Gemini Advanced web
Use APIs for speed and reliability. Use browser for subscription-only features or comparison.
User: "Use browser to compare how ChatGPT writes about statins"
Claude: I'll get ChatGPT's perspective using browser automation.
[Uses browser_navigate to https://chat.openai.com]
[Uses browser_snapshot to verify page state]
[Uses browser_type to enter: "Write a patient-friendly explanation of how statins work"]
[Uses browser_click to send]
[Uses browser_wait_for to wait for response]
[Uses browser_snapshot to capture response]
Here's what ChatGPT wrote:
[Response text]
Compared to my approach:
[Claude's version]
Key differences:
- ChatGPT emphasized X while I focused on Y
- Tone: ChatGPT more conversational, mine more clinical
- Length: Similar word count
Browser automation gives you access to your paid subscriptions programmatically, complementing the API-based models in your arsenal.