--- name: arxiv description: Fetch an arXiv paper PDF and LaTeX source into .research/references using the bundled fetch script. --- # arXiv Fetch Fetch arXiv papers and their LaTeX source into the project's reference store. ## Layout Store every paper under `.research/references//`: - `.pdf` — the published PDF. - `source/` — the extracted e-print source (`.tex`, figures, `.bbl` inputs). - `meta.json` — id, URL, fetched date, and which artifacts succeeded. Normalize `` as the full arXiv identifier with version when given (for example `2301.12345v2` or `hep-th/9901001v1`), stripping any URL prefix. ## Workflow 1. Run the bundled script from the project root: `sh /skills/arxiv/scripts/fetch-arxiv.sh [--pdf-only | --source-only]` 2. Do not invent the PDF or source contents. If either download fails, report which artifact failed and keep whatever succeeded. 3. Read the extracted `.tex` files for citations, notation, or proofs; cite the arXiv id in any summary. ## Script behavior - Uses `curl` with retries and fails on HTTP errors. - PDF comes from `https://arxiv.org/pdf/`. - Source tarball comes from `https://arxiv.org/e-print/` and is extracted with `tar` into `source/`. Some e-prints are single files rather than tarballs; the script keeps those as `source/eprint-file`. - Creates `.research/references/` when missing. Never writes outside it.