devpilot-learnlisted

# Learn Generate a clean, reader-mode HTML digest from a single content source. The digest captures the article's core argument, key points, and important visuals in a format that's easy to scan and share. ## Workflow ### 1. Identify the input source The user will provide one of: - **URL** — a web page or online article - **PDF** — a local file path to a .pdf - **Word doc** — a local file path to a .docx - **Text file** — a local file path to .txt, .md, or similar - **Pasted text** — content directly in the message Detect which type it is and proceed accordingly. ### 2. Fetch the content **For URLs:** Use `WebFetch` to retrieve the page. If the page content is too short or looks like a paywall/login wall, tell the user and ask if they have another way to access it. **For PDFs:** Use the `Read` tool with the file path. For large PDFs (>10 pages), start with the first 20 pages, summarize, then continue if needed. **For Word docs (.docx):** Use `Bash` to convert with `textutil -convert txt <file> -stdout` (macOS) or `pandoc <file> -t plain` as a fallback. If neither is available, try reading the file directly and extracting what you can. **For text files / pasted text:** Read directly. ### 3. Determine output language - If the user explicitly specifies a language (e.g., "summarize in English", "用中文总结"), use that language. - If no language is specified, default to the language of the original content. - If the content is mixed-language, default to the dominant lang