Reviews academic papers as an adverse referee with page-anchored objections and derived verdicts.
npx skills add https://github.com/BTreeMap/SKILLs --skill peer-review-btreemap
Please help me install this Agent Skill. Skill: peer-review Source: https://github.com/BTreeMap/SKILLs/tree/main/peer-review Command: npx skills add https://github.com/BTreeMap/SKILLs --skill peer-review-btreemap
High-intent search queries and answers about installing and using this skill.
Initialize a session with the paper's title and date, ingest text extracted via the read-pdf skill, then note claims and objections in JSON batches. Every objection must quote the paper verbatim or name a specific omission, and the check command derives the final recommendation.
Novelty objections require corpus keys from a linked lit-review session whose records predate the paper. Claims of being first or state-of-the-art are refused by the gate unless they cite dated prior work found through an actual literature search.
Three levels: lite walks only the claims and limitations banks with abstract-level reading, full walks all five banks with a literature corpus, and ultra adds recomputation of reported numbers and forward snowballing from every prior-work citation.
No. Every objection must either quote the paper verbatim (resolved to a page at 0.85 similarity or better) or name a missing item with its expected location. Objections that fail anchoring are rejected and cannot enter the report.
The review only covers what the extracted text and linked corpus support; without network access the novelty bank is unassessed, and papers with no reachable text run at lite only. It also cannot judge correctness of results beyond internal consistency and recomputation.
rbudiharsoExtract text, tables, and images from PDFs and scanned documents using pymupdf or marker-pdf.
pedrohcgsGenerate a structured response-to-referees document mapping referee comments to manuscript revisions.
TylerGarlickVerify claims against provided sources and flag unsupported citations.
001TMFIngest wet-lab readouts and calibrate in-silico predictions against lab outcomes.
