Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when drafting or polishing the prose of a Comparative Political Studies (CPS) manuscript so it reaches comparativists, fits the 11,000-word limit, and follows CPS house formatting. Improves writing and formatting; it does not change the analysis.
.claude/skills/brycewang-stanford-cps-writing-style/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 41% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 10% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 7% | 0% |
A CPS paper must read for the comparative-politics community: a comparativist of another region should follow the argument, see the leverage, and grasp the portable claim. Write argument-first, keep the prose tight against the word cap, and follow CPS/SAGE house formatting (APA-style author-date references, unstructured 150-word abstract).
Lead with the comparative puzzle, then the stake, then why it is hard to identify, then the setting & design, then the headline finding with mechanism, then the portable contribution, then a brief roadmap. By the end of the introduction the reader knows the question, the argument, the evidence, and why it matters to comparative politics — not just to one country's specialists.
literature tour or lead with the estimator.
past your subfield within comparative politics.
| Item | CPS norm (检索于 2026-06;以官网为准) | |------|------| | Length | Articles max 11,000 words; references, tables, figures excluded from the count | | Abstract | Unstructured, ~150 words, between title and body | | References | APA-style author-date (SAGE house) — every in-text cite has a list entry and vice versa | | Manuscript | Double-spaced, 1" margins, 12-pt (Times New Roman preferred) | | Anonymity | Remove identifying info from the main file; put it on a separate title page |
> Length and house style change — confirm on the SAGE author-instructions page; treat figures above as > the working norm.
toward the limit) — but keep the story in the main text.
【Intro arc】puzzle → stake → identification problem → design → finding+mechanism → portable contribution → roadmap
【Travels?】contribution framed for a comparativist of another region [Y/N]
【Length】words / 11,000 (refs/tables/figures excluded)
【Abstract】unstructured, ~150 words [Y/N]
【References】APA author-date, in-text=list [Y/N]
【Anonymized】identifying info off the main file [Y/N]
【Next】cps-transparency-and-data../../resources/worked-examples/01-introduction.md — before→after CPS-style introduction../../resources/official-source-map.md — CPS length, abstract, and style facts with sourcing| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | pass→pass | 15,958 | 15,062 | -6% | 1 | 1 | 0% | 2,159 | 2,306 | +7% | 0 | 0 | — |
case-07 | fail→pass | 13,557 | 14,451 | +7% | 1 | 1 | 0% | 1,489 | 2,458 | +65% | 0 | 0 | — |
case-01 | fail→pass | 16,054 | 13,339 | -17% | 1 | 1 | 0% | 1,690 | 2,389 | +41% | 0 | 0 | — |
case-02 | fail→pass | 26,061 | 18,301 | -30% | 1 | 1 | 0% | 4,061 | 3,197 | -21% | 0 | 0 | — |
case-03 | fail→pass | 26,337 | 16,793 | -36% | 1 | 1 | 0% | 3,254 | 3,582 | +10% | 0 | 0 | — |
case-04 | pass→fail | 32,373 | 33,349 | +3% | 1 | 1 | 0% | 5,551 | 6,082 | +10% | 0 | 0 | — |
case-05 | pass→fail | 26,433 | 27,280 | +3% | 1 | 1 | 0% | 3,100 | 4,486 | +45% | 0 | 0 | — |
case-06 | pass→pass | 15,539 | 20,665 | +33% | 1 | 1 | 0% | 2,599 | 3,551 | +37% | 0 | 0 | — |
case-08 | pass→pass | 16,840 | 22,754 | +35% | 1 | 1 | 0% | 1,917 | 2,775 | +45% | 0 | 0 | — |
case-09 | pass→pass | 19,996 | 17,987 | -10% | 1 | 1 | 0% | 2,230 | 2,848 | +28% | 0 | 0 | — |
case-10 | fail→pass | 21,483 | 11,925 | -44% | 1 | 1 | 0% | 2,023 | 2,171 | +7% | 0 | 0 | — |
case-11 | pass→pass | 13,779 | 16,733 | +21% | 1 | 1 | 0% | 1,894 | 2,908 | +54% | 0 | 0 | — |
case-12 | pass→pass | 28,700 | 24,150 | -16% | 1 | 1 | 0% | 2,763 | 3,446 | +25% | 0 | 0 | — |
case-13 | pass→pass | 22,962 | 13,066 | -43% | 1 | 1 | 0% | 2,121 | 3,159 | +49% | 0 | 0 | — |
case-14 | pass→pass | 22,303 | 19,792 | -11% | 1 | 1 | 0% | 2,171 | 2,765 | +27% | 0 | 0 | — |
case-15 | fail→pass | 11,768 | 3,413 | -71% | 1 | 1 | 0% | 957 | 1,578 | +65% | 0 | 0 | — |
case-16 | fail→pass | 10,212 | 12,099 | +18% | 1 | 1 | 0% | 1,200 | 2,629 | +119% | 0 | 0 | — |
case-17 | fail→pass | 18,021 | 7,902 | -56% | 1 | 1 | 0% | 1,694 | 1,428 | -16% | 0 | 0 | — |
case-18 | fail→pass | 15,843 | 16,273 | +3% | 1 | 1 | 0% | 2,083 | 2,834 | +36% | 0 | 0 | — |
case-20 | pass→pass | 15,119 | 9,416 | -38% | 1 | 1 | 0% | 1,372 | 1,669 | +22% | 0 | 0 | — |
case-21 | pass→pass | 9,701 | 3,146 | -68% | 1 | 1 | 0% | 759 | 1,414 | +86% | 0 | 0 | — |
case-22 | pass→pass | 15,648 | 13,882 | -11% | 1 | 1 | 0% | 2,054 | 2,933 | +43% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +32 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.