自动追踪 AI 研究方向与论文

Build and maintain a per-direction LLM Wiki (Obsidian-style knowledge base) of AI research papers. Sets up a fresh data repo, registers a research direction, cold-start backfills ~6 months of arXiv + surveys, then does daily incremental updates that fetch new papers, filter by relevance, extract full text + figures, compile cross-linked Chinese wiki pages, and emit a daily briefing (简报). Use this skill whenever the user wants to set up or initialize a paper/literature knowledge base, start tracking an AI research direction or topic, build a research wiki, fetch or pull "today's" / latest arXiv papers for a topic, run a daily/scheduled paper update, generate a research briefing or digest, quickly read/summarize a single paper (by arXiv id/URL or local PDF) to pull out its key info, query an existing paper knowledge base, or health-check (lint) a vault — even if they don't say "research-wiki" by name. Supports multiple independent directions via a registry; each direction is one self-contained vault.