reading-paperslisted

Look up, resolve, and read academic papers when the user sends a link, DOI, title, author name, or vague description, across marketing (Marketing Science, JMR, JM, JCR), economics (QJE, AER, ECMA, JPE, ReStud, AEJs, AER:Insights), psychology (Psych Science), and CS (ICLR/ICML/NeurIPS, JMLR, ACL). Preserves LaTeX equations; integrates the user's Zotero library and an optional ~/ocr-examples HPC OCR pipeline. TRIGGER whenever the user asks to read, summarize, check, find, or cite a paper, asks "what does X argue", pastes a paper URL, or asks what an author has written. For a whole literature on a topic, use litreview instead.
ericluo04/claude-academic-workflow · ★ 5 · AI & Automation · score 67

Install: claude install-skill ericluo04/claude-academic-workflow

# Reading papers Get from a fuzzy reference to clean full text with equations intact, cheaply, without wasting context on dead ends. The intelligence is yours; the script is a tool. Don't blindly trust its first answer on hard cases (common author names, paywalled venues). Verify, as flagged below. ## The command ```bash ~/.claude/skills/reading-papers/scripts/paper.py <cmd> "<query>" [flags] ``` | cmd | does | |---|---| | `resolve` | fuzzy ref → canonical record (DOI, arXiv id, venue, year, OA status, best free URL, free AEA appendices) | | `get` | resolve, then fetch the cleanest full text available | | `search` | topic/keyword search across Semantic Scholar + OpenAlex + arXiv, deduped and merged | | `author` | works by a person, newest first, with disambiguation | | `cites` | citation graph: cited-by and references; `--contexts` adds the citing sentences | Flags: `--venue "…"` · `--since/--until YEAR` · `-n N` · `--json` · `--save` · `--raw` · `--list-sections` · `--section "<substr>"` · `--affiliation "…"` · `--orcid "…"` · `--contexts` Keys (optional, all free) live in `~/.claude/secrets/scholar.env`; the script auto-loads them. ## How resolution works, and why order matters `resolve` is cheap and prints whether a free copy exists and where. Always resolve before web-searching or guessing URLs. The routing is cost- and Cloudflare-aware: - An arXiv id or arxiv link → arXiv metadata (free) + OpenAlex if a DOI exists. - A DOI → OpenAlex (1 credit) + Unpaywall + NBE