Review Paper Agent Skill

Review-Paper Skill

Scaffold and draft a literature review — narrative, scoping (PRISMA-ScR), or systematic

(PRISMA 2020) — for medical / medical-AI research. This skill builds the structure, the

required scope/non-overlap framing, the summary-table stubs, and the reporting-guideline

wiring, then hands off to the existing QC skills. It is the review-article counterpart to

write-paper (which targets original research); for *reviewing* someone else's review

article, use /peer-review or /self-review (the RV1-RV9 probes). The structure follows

established review-writing conventions; it is not derived from, and does not reproduce, any

specific published review.

Anti-Hallucination

• Never invent citations. Every citekey must resolve to the project's verified

_src/refs.bib (produced by /search-lit → /lit-sync → /verify-refs). If a claim

needs a reference that is not yet in the library, leave a [NEEDS-REF: claim] marker and

route it to /search-lit; do not fabricate a DOI, author, year, or citekey.

• Never invent data. Summary-table cells (study, year, metric, finding) are filled only

from sources the user supplies or that are verified; an unknown cell stays a placeholder.

• No recommendation-grade language without standing. For a scoping review especially,

the output maps the evidence — it does not issue clinical recommendations.

• Quality gate before hand-off: the draft is not "done" until /self-review reports 0

fatal findings and /verify-refs reports 0 FABRICATED / MISMATCH and no placeholder

citations remain.

Step 0 — Format + spine axis (the structure-determining choice)

• Confirm format with the user: narrative (SANRA) | scoping (PRISMA-ScR + JBI) |

systematic (PRISMA 2020). This decides the reporting guideline and the registration path.

• Choose the spine axis — the single most consequential decision: organize the body by

modality (e.g. 2D → 3D), by task (generation / QA / deployment), or by

lifecycle stage. Every body section then follows this one axis; mixing axes is the

most common structural failure.

• Require a scope statement + non-overlap boundary against prior/adjacent reviews — this

pre-empts the reviewer's first question, "why another review on this?" (user-approval

checkpoint: confirm the boundary with the user before scaffolding).

Step 1 — Scaffold the 7-part macro skeleton

Load ${CLAUDE_SKILL_DIR}/references/macro_skeleton.md and instantiate:

• Abstract — structured for scoping/systematic; a 4-5 move version for narrative.
• Introduction — clinical motivation → technology → **scope + non-overlap block

(required field)** → "this review…".

• Background / technical principles — tight; cite once, do not re-survey the field.
• Thematic body by spine axis — each section ends with a summary table (stub

generated to match the type, Step 2).

• Frontiers / what is advancing.
• Challenges / discussion — include an evaluation-metrics critique subsection (a

required quality signal: how the field measures itself, and where those metrics mislead).

• Conclusion — measured; no recommendation-grade language for a scoping review.

Step 2 — Summary-table stub (matched to type)

• Narrative / scoping: study | year | [spine-axis value] | method | key finding.
• Systematic: PRISMA flow + a study-characteristics table + an extraction table.

The stub ships with column headers and one placeholder row; rows are filled only from

verified sources (see Anti-Hallucination).

Step 3 — Reporting + registration wiring

• Scoping → PRISMA-ScR (+ JBI charting) + OSF registration.
• Narrative → SANRA (a 6-item appraisal aid, not a reporting checklist — do not

over-enforce it).

• Systematic → PRISMA 2020 (+ PROSPERO registration).
• If /check-reporting does not yet carry the chosen checklist (e.g. PRISMA-ScR), track a

manual gap table and flag it for the user rather than silently skipping the item.

Step 4 — QC hand-off

Run the standard manuscript QC chain, which this skill is designed to feed:

• /self-review — the RV1-RV9 narrative-review probes auto-activate for a review article.
• /check-reporting — the chosen guideline (SANRA / PRISMA-ScR / PRISMA 2020).
• /verify-refs — every citation resolves; 0 FABRICATED / MISMATCH.
• /humanize — AI-pattern density below threshold.
• /academic-aio — discoverability pass (optional).

Convergence gate: self-review fatal = 0; verify-refs FABRICATED/MISMATCH = 0; no

[NEEDS-REF] / [@NEW:]-style placeholder citations remain; humanize density < 2.0.

Guards

• Citations resolve to _src/refs.bib only; never invent citekeys (see Anti-Hallucination).
• Proportionate self-citation; declare an intellectual conflict of interest when an author

has contributed to the area being reviewed (per intellectual-coi).

• Write only inside the manuscript directory.

references/

• macro_skeleton.md — the 7-part template and the table/figure plan per review type.