paper-reviewlisted

# Paper Review ## Overview A pre-submission **novelty + reviewer pass** that grounds every judgment in a *retrieved* paper. Naive LLM review asserts novelty from memory and invents citations; this skill retrieves real prior work, quotes it, and cross-checks the draft's bibliography — then surfaces (without adjudicating) any overlap with a lab or co-author line that has to be disclosed. **Core principle:** a novelty claim is only as good as the paper it is checked against. Every verdict cites a record retrieved this run with a verbatim snippet, or it is marked **unverified** — the analogue of `techne:paper`'s "never hand-type a result number". ## When to use - A first-author draft is approaching submission and needs a distinctness + related-work check. - "Review `<name>` for novelty", "what am I missing", "is C2 actually new". - NOT for copy-editing prose; NOT for a paper with no stated contributions yet (scaffold first with `techne:paper`). ## Inputs - `<name>` — the paper directory under `papers/` (e.g. `velocity-fl-systems`). - `--from <repo>` (optional) — the source code repo; grounds quantitative claims against its harvest output. ## Config Read `## paper-review` from `<repo>/.claude/skill-context.md`: `lab_line` (overlap source of truth, default `LINEAGE.md`), `retrieval` (default OpenAlex + web; Semantic Scholar only if `S2_API_KEY` is set), `mailto` (OpenAlex polite pool, default `git config user.email`), `report` (default `papers/<name>/novelty-review.md`