ContentForge

Problem

Chinese creators often convert articles, podcasts, finance news, community posts, or videos into podcasts, PPTs, mind maps, and analysis notes, but manual summarizing, reformatting, Feishu sync, and IMA ingestion are fragmented. ContentForge turns that into a repeatable SOP: match a scenario, extract or organize source content, generate with NotebookLM or a local LLM fallback, then write results to Feishu docs, bitables, or IMA knowledge bases.

How It Works

Scenario routing selects the right document when the input is a WeChat article URL, Xiaoyuzhou URL, post text, Caixin URL, or video link. The shared workflow checks notebooklm status, uses Playwright for text upload and generation, and degrades to local generation when needed. Outputs separate facts, quotes, *(inference)*, and *(AI suggestion)*, and include a confidence field.

Boundaries

This skill depends on Chinese content extraction, Feishu/IMA integration tools, and browser automation. Paywalls, restricted networks, or incomplete fetches use fallback references. NotebookLM audio, video, and report generation are subject to account-level daily quotas, while mind maps and some notebook artifacts can reset by creating a new notebook.