paper-reviewlisted

# Paper Review A six-phase simulated peer-review panel. It takes the author's **own draft** and returns a structured referee report plus a scorecard verdict — the kind of feedback a serious reviewer would give, calibrated to how finished the draft is. ``` intake (+type) ─▶ 5 lenses ─▶ rebuttal round ─▶ moderator ─▶ editor letter + scorecard ``` This is pure reasoning — no scripts. **You are one Claude playing all six roles in sequence within a single turn** — the "5 lenses" framing is conceptual, not a multi-agent runtime. The per-phase JSON is internal scratch; by default you deliver only the **review letter** and the **scorecard table**. The full pipeline, the five lenses, the **paper-type → standard mapping** (CONSORT, STROBE, PRISMA, COREQ, validity typology, Toulmin, NIH), the score/priority calibration, the draft-level tones, the verdict thresholds, the transparency & ethics flags, and every verbatim role prompt are in `references/review.md` — **load it before reviewing.** ## How to run it 1. **Default = review immediately. Don't gate behind a confirmation menu.** Once you have the draft, just review it — infer the **draft stage** from the message (any signal word → that stage; otherwise default to `working`), default to all five lenses and the user's language. **Don't ask the user to confirm things they already said.** Append ONE line after the verdict so they can correct the one thing that matters: *"I reviewed this as a **working draft** — if