AgentRender Web Capture

Web pages are often inaccessible to agents due to dynamic rendering or lack of direct browser access. This skill provides a reliable bridge to capture public web content as high-fidelity visual evidence, printable archives, or structured data. It eliminates the friction of manual "save as" workflows and provides cryptographic proof of content via render receipts. ## What it does - **Visual capture** creates full-page screenshots for UI audits, layout verification, and visual regression checks. - **Portable archiving** generates PDF versions of invoices, documentation, or long-form articles for sharing and storage. - **Data extraction** parses page titles, text bodies, and link sets into structured JSON for immediate LLM reasoning. - **Verifiable receipts** provides final URLs, timestamps, and content hashes to ensure data provenance and auditability. ## How it works 1. Use `screenshot` when you need to inspect the visual layout or detect UI bugs on a public URL. 2. Call `pdf` to archive content into a durable format suitable for printing or document management systems. 3. Execute `extract` to pull text and link data for analysis without consuming excessive context window space with image data. 4. Reference the **render receipt** in your output to confirm the exact capture time and final URL after any redirects. ## Frameworks & tools This skill utilizes the AgentRender MCP server. It supports both remote hosted MCP via HTTP and local execution using `uvx`. ## Why this beats prompting it yourself Standard agents often hallucinate web content or fail when pages require JavaScript execution. This skill uses a dedicated rendering engine to ensure the agent sees exactly what a browser sees, backed by receipts that prevent "hallucinated" page data. ## Use cases - Capture visual proof of site deployments or UI bugs for bug reports. - Extract structured text from Wikipedia or documentation for RAG pipelines. - Save copies of online receipts and invoices directly to a project folder. - Verify site redirects and metadata for SEO or security audits. ## Known limitations Does not support pages requiring authentication, paywalls, or local network access. Limited to public http/https schemes. Requires an AgentRender API key.