PubMed Literature Search

Problem

When doing biomedical literature review, PubMed indexes over 30M records, but PMC full text covers only about 4M open-access articles. Harvesting PMC alone misses most non-OA papers, while checking full text manually is blocked by paywalls and ID mapping. This is especially visible when results span multiple years, journals, and MeSH topics, and manual summarization or OA labeling is error-prone.

How It Works

The skill splits retrieval, summarization, and OA detection into a stable workflow, fixing the order of NCBI E-utilities calls and reducing manual URL assembly and ID guessing:
- Runs PubMed queries via esearch, supporting tiab, mesh, journal, dp, author, and other full syntax;
- Uses esummary to inspect pmcid for open-access status;
- Fetches structured abstracts with efetch;
- Retrieves full-text XML via OAI-PMH when a PMCID exists;
- Forces each result to be labeled [OA] or [Non-OA], with PMID, PMCID, and batch OA statistics.

It is built for API-friendly use: no API key is required, and built-in delays help stay within the public rate limit.

Boundaries

Non-OA articles return abstracts and metadata only. Large batches during NCBI peak hours may be limited, and older pre-1975 records may have incomplete metadata. For full text, rely only on [OA] items or use institutional access, publishers, and DOI links. It is better suited for literature lists, abstract comparison, and OA filtering than for formal citation validation, full-text OCR, or database authorization compliance.