// Fetch full-text articles and metadata from PubMed/PMC. Use when a PDF is missing from literature/ and you need the article text or abstract to verify a claim. Full text is only available for open-access PMC articles (~6 million).
This skill should be used when creating, editing, or validating GO-CAM (Gene Ontology Causal Activity Model) models. It provides comprehensive annotation guidelines for molecular functions, biological processes, cellular components, and causal relationships following GO Consortium standards.
Search GO bioentities (genes, gene products, complexes) and their GO annotations directly from the GO database. Use to resolve MGI/MOD/ComplexPortal IDs, cross-check GO annotations independently of gocam search, or find MOD-specific gene product IDs needed in Noctua models.
Independent blind evaluation of GO-CAM claims documents using Claude Explore subagents to read the PDFs. Reads expert_validation_claims.md and verifies every claim against primary PDFs from scratch, producing a structured evaluation report. Use after gocam-claims-pipeline has generated its output. Always run in a fresh session — never in the same session as the pipeline. For a Gemini-powered alternative, see /blind-review-gemini.
Independent blind evaluation of GO-CAM claims documents using Gemini. One Gemini instance per primary PDF reads the paper directly and returns PASS/FLAG/FAIL verdicts on every claim citing it. Use after gocam-claims-pipeline has generated its output. Always run in a fresh session — never in the same session as the pipeline. For a Claude-engine alternative, see /blind-review-claude.
Generate a .drawio pathway diagram from a GO-CAM claims document. Use after completing the annotation pipeline (Phase 7) to produce agent-output/pathway_sketch.drawio. Reads shape styles and the Shape Palette page from workflow/network_template_draw.drawio.
Launches two parallel Gemini instances per paper to review GO-CAM claims from two complementary perspectives — a peer reviewer (scientific accuracy) and a GO-CAM curator (annotation correctness). Each receives the full claim in claims_template format and responds as an expert with OK/WRONG/UNCERTAIN plus a corrected claim where needed. Collects both outputs into peer_review_claims.md alongside the original claim.
| name | pubmed |
| description | Fetch full-text articles and metadata from PubMed/PMC. Use when a PDF is missing from literature/ and you need the article text or abstract to verify a claim. Full text is only available for open-access PMC articles (~6 million). |
Use when a PDF is not available in literature/ and you need to verify a claim against the actual paper.
Step 1 — Convert PMID to PMCID:
mcp__claude_ai_PubMed__convert_article_ids(pmids=["12345678"], target_id_type="pmcid")
Step 2 — Fetch full text:
mcp__claude_ai_PubMed__get_full_text_article(pmc_ids=["PMC1234567"])
Returns the full article text. ~6 million open-access articles are available. If the article is paywalled, this returns nothing — fall back to abstract.
Context budget warning: Full-text articles can be 20,000–50,000 words. Do not dump the entire text into your response. Extract only the sections you need (Results, Methods for the relevant figures) and discard the rest. If delegating to a subagent, instruct it to return only the specific passages that answer your verification question.
mcp__claude_ai_PubMed__get_article_metadata(pmids=["12345678", "87654321"])
Returns title, authors, journal, year, abstract. Accepts arrays — batch multiple PMIDs in one call. Use when full text is unavailable. Mark any claim verified only from abstract as "PDF not available — abstract only" in errors_report.md.
mcp__claude_ai_PubMed__search_articles(query="synaptic vesicle exocytosis Munc18 mouse", max_results=20)