Scaffold and draft medical/AI literature reviews (narrative, scoping PRISMA-ScR, or systematic). Asks for the spine axis, builds a 7-part skeleton with a required Intro scope/non-overlap block, a summary-table stub, an evaluation-metrics critique subsection, and reporting-guideline wiring. Reuses the self-review RV1-RV9 narrative-review probes for QC. Does not invent citations.
SkillScore87Maintained
Automatic
◆Research pack
Last verified 8/24/2026
What it changes
Passed 19 of 22 test cases with the skill, and 12 of 22 without it.
Without skill57%
With skill87%
Checks it passed that the model alone missed
Flags the manual gap table for user attention.
Inserts a [NEEDS-REF: claim] marker for the unverified assertion.
Includes all seven macro skeleton sections in the manuscript draft.
and 6 more in the full report
8 fixed12 passed either way2 still failing1 broke
Pass-rate lift
+30 pp
Tokens
+28%
Agent turns
0%
Response time
-22%
Model gemini-3.6-flashMeasured 8/24/2026v1 · latest is v220 of 23 cases passed with skill · 22 comparable
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
A newer attempt on gemini-3.6-flash on 10/3/2026 did not produce a publishable comparison (not comparable). The snapshot above is the previously measured run.
See it in action
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 7
Prompt
In our literature review draft on AI in radiology, we state that vision transformers outperform CNNs on large pretraining datasets, but we do not yet have a verified bibtex citation in _src/refs.bib. Should we construct a citation like (Dosovitskiy et al., 2021) or invent a citekey [@dosovitskiy2021]?
What was checked, without and with the skill
Checked
Without skill
With skill
Refuses to fabricate citation strings or citekeys missing from _src/refs.bib.
Passed
Passed
Inserts a [NEEDS-REF: claim] marker for the unverified assertion.
Missed
Passed
case-10 · gemini-3.6-flash · 8/24/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md87 lines2 files
Review-Paper Skill
The review-article counterpart to write-paper (original research). For reviewing someone
else's review article, use /peer-review or /self-review (the RV1-RV9 probes) instead.
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 macro skeleton for the format
Load ${CLAUDE_SKILL_DIR}/references/macro_skeleton.md and write the section for the chosen
format to manuscript.md in the manuscript directory, with the summary-table stubs (Step 2) and
that section's figure/table legend plan. The structure depends on the format:
Narrative — the 7-part skeleton below.
Scoping / systematic — IMRaD with one Methods and one Results slot per PRISMA-ScR or
PRISMA 2020 item (the reference lists them by item number). Do not use the 7-part skeleton
for these: it has no place for eligibility criteria, search, selection, charting/extraction,
risk of bias or synthesis methods, nor for study-selection results. The thematic body by
spine axis goes inside Results → synthesis; frontiers and the evaluation-metrics critique go
inside Discussion.
The narrative 7-part skeleton:
Abstract — a 4-5 move version (structured for scoping/systematic).