lit-review

What problem does it solve? Writing a literature review from memory or unverified search snippets produces citations that cannot be traced, exclusions without reasons, and claims that overstate how deeply sources were read. This Skill enforces a logged, criteria-first review process where every citation resolves to a retrieved paper record. ## Core Features & Use Cases - Logged multi-source search: Runs queries against OpenAlex, arXiv, and Crossref through a bundled script that records every query, count, and snowball round, with deduplication across sources. - Criteria-gated screening: Inclusion and exclusion criteria must exist before the first search; every exclusion carries a reason, and bulk regex rules screen whole clusters of candidates at once. - Read-level honesty and verification: Each claim carries the read level of its source (abstract vs full-text), and citation checks plus DOI verification run before the report is delivered. - Use Case: Ask for a systematic review of retrieval-augmented generation evaluation methods; the Skill fixes criteria, searches three indexes, snowballs citations, screens with logged reasons, and delivers a PRISMA-style report where every [n] citation traces to a corpus record. ## Quick Start Ask the agent to run a full literature review on your research question, for example: "Run a full literature review on how retrieval-augmented generation affects factual accuracy in question answering."