Paper Analyst

Reads academic papers from PDFs or text and produces structured analysis in Chinese by default. Built around three reference files that enforce anti-hallucination rules: every claim gets tagged as either directly stated in the paper or inferred by the model, uncertain fields are marked explicitly, and it won't fabricate missing metadata. Handles degraded PDFs by noting what's unreadable instead of filling gaps. The presentation modes generate slide outlines that compress findings into talk-ready formats, then automatically hand off to a separate pptx skill for rendering. Useful when you're prepping for lab meetings or need to verify what a paper actually claims versus what people say it claims.