- Home
- Skills
- Workflow Automation
- Agent Playground
More screenshots
Works with the AI Agents you already use
Agent Playground
Playwright-powered toolkit for navigating JS-heavy sites and extracting structured data from GitHub and social platforms.
Free
Agent Playground
Example session with this skill installed
Search GitHub for the top 3 trending AI agent repositories sorted by stars and extract the results.
- Read your context and instructions
- Compiled the agent playground
I have performed a GitHub search for "ai agent" repositories. The browser successfully navigated the dynamic results page, sorted by star count, and extracted the following metadata. ## GitHub Search Results: AI Agents | Repository | Stars | Description | URL | | :--- | :--- | :--- | :--- | | **significant-gravitas/AutoGPT** | 165,000+ | An experimental open-source attempt to make GPT-4 fully autonomous. | `https://github.com/significant-gravitas/AutoGPT` | | **assafelovic/gpt-researcher** | 12,000+ | LLM based autonomous agent that conducts deep local and web research. | `https://github.com/assafelovic/gpt-researcher` | | **transformer-agents/transformer-agents** | 8,500+ | A library for building and deploying agentic workflows with Transformers. | `https://github.com/transformer-agents/transformer-agents` | **Next steps** 1. Run `python scripts/run.py explore --url [URL]` on any of these repositories to extract detailed README content. 2. Use the `status --platform github` command to verify your session is still active for deeper API interaction. 3. Capture a screenshot of the AutoGPT star history using the `--capture` flag for visual analysis.
Connects securely to your tools. The creator never sees your data.
About this skill
JS-heavy websites and SPA feeds often break standard scrapers, returning empty divs or bot-detection screens. This skill provides a Playwright-powered toolkit to navigate dynamic sites, extract clean Markdown, and interact with specific platforms like GitHub, Bilibili, and Zhihu. ## What it does - **Dynamic Exploration** executes multi-step scrolling to trigger lazy-loading and extracts clean text and links. - **Visual Auditing** captures full-page screenshots to provide agents with visual context of complex layouts. - **Platform Integration** uses specialized handlers for GitHub search and repository metadata retrieval. - **Session Management** maintains persistent cookies for authenticated platforms to avoid repeated logins. - **Multi-Browser Support** utilizes system Chrome with an automatic fallback to Microsoft Edge. ## How it works 1. **Explore Mode** uses `explore --url` to crawl dynamic pages and convert them into agent-readable Markdown. 2. **Search Mode** invokes platform-specific commands like `github-search` to fetch filtered, sorted repository data. 3. **Status Check** monitors session health across active platforms using the `status` command. 4. **Visual Debugging** uses the `--visible` flag to launch a headed browser for manual intervention or monitoring. ## Frameworks & tools Built with Python and Playwright. It targets Chromium-based browsers (Chrome, Edge) and integrates with LLMs via the Model Context Protocol (MCP) or a standalone CLI. ## Why this beats prompting it yourself Generic agents often fail to wait for JS hydration or handle infinite scroll, leading to "no content found" errors. This skill encapsulates low-level browser automation logic, ensuring the agent receives structured data instead of raw, messy HTML. ## Use cases - Monitor GitHub trending repositories for specific tech stacks. - Extract clean content from JS-heavy social feeds like Bilibili or Zhihu. - Debug web layouts by reviewing automated full-page screenshots. - Automate search and discovery across developer-centric platforms. ## Known limitations Requires local installation of Python dependencies and Chromium-based browsers. Cannot bypass CAPTCHAs or anti-bot gates; manual login via visible mode is required for blocked sessions.
Requires Python and local Chrome/Edge. Cannot bypass CAPTCHAs or bot-protection; requires manual intervention for login challenges.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 5 days ago
- Free with an Agensi account
Permissions
Uses Playwright with local Chrome/Edge to browse dynamic pages and extract Markdown. Optional platform sessions are saved locally under storage/cookies/.