Literature Mining Expert

Problem: The bottleneck is synthesis, not collection

After searching, the hard part is usually not finding more papers, but extracting consistent findings from a pile of PDFs, comparing methods, mapping claims, and deciding which gaps are real. literature-mining targets the reading and synthesis stage: it does not perform broad literature discovery, but turns user-supplied papers, abstracts, PDF text, or curated lists into reusable research artifacts.

How it works: from single-paper notes to a literature matrix

The skill acts as a methodology toolkit. Core outputs include structured reading notes, a literature matrix, gap identification, and a review framework. A typical workflow is:
- Define a search strategy using PICO when the scope needs tightening, even if retrieval is handled elsewhere;
- Record PRISMA-style screening so inclusion and exclusion decisions remain auditable;
- Build reading notes for core papers, covering question, method, data, conclusions, and limitations;
- Compare multiple papers in a literature matrix across methods, metrics, settings, and claims;
- Identify gaps from the matrix and test whether they are actionable innovation points;
- Grade evidence quality so weak and strong evidence are not mixed together in the review.

Boundaries: read-only analysis, no fabrication

This skill is suitable for close reading, cross-paper comparison, review scaffolding, and evidence grading. It is not a replacement for web_search or dedicated retrieval skills. If network access or external search tools are unavailable, it should continue only with the materials the user has already provided and state that limitation. It must not invent papers, authors, DOIs, or references, and every output should be treated as AI-assisted analysis that needs verification against the original sources.