Use when designing tables and figures for a Comparative Political Studies (CPS) manuscript so each exhibit is self-contained, comparative-legible, and SAGE-ready. Improves the exhibits; it does not run the analysis or set the data policy.
SkillScore91Maintained
Automatic
◆Research pack
Last verified 9/1/2026
What it changes
Passed 21 of 23 test cases with the skill, and 15 of 23 without it.
Without skill65%
With skill91%
Checks it passed that the model alone missed
The response includes [Y/N] in the 【Production】 section.
The response includes [Y/N] in the 【Self-contained?】 section.
The response specifies cps-writing-style in the 【Next】 section.
and 4 more in the full report
6 fixed15 passed either way2 still failing0 broke
Pass-rate lift
+26 pp
Tokens
+25%
Agent turns
0%
Response time
-27%
Model gemini-3.6-flashMeasured 9/1/2026v121 of 23 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 6 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 6
Prompt
We are using Stata integrated with automated tool workflows to programmatically export our primary event-study plot from a fitted model result without manual data transcription. Which tool command should be called to generate this plot directly from the result object?
What was checked, without and with the skill
Checked
Without skill
With skill
The response names enhanced_event_study_plot for Stata figure generation from results.
Missed
Passed
case-16 · gemini-3.6-flash · 9/1/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md93 linessingle file
Tables & Figures (cps-tables-figures)
CPS exhibits must carry the comparative argument on their own. A reviewer should grasp the cross-case or
over-time story from the figure and its caption without hunting through the text. Because references,
tables, and figures do not count toward the CPS word limit, exhibits are where you spend space —
but only if each one earns its place.
When to trigger
Building or revising the main-text tables and figures
A reviewer found exhibits unreadable, over-stuffed, or not self-contained
Deciding which exhibits belong in the paper vs. the supplementary/online appendix
Preparing figures for SAGE production (color, resolution, grayscale legibility)
Exhibit priorities (in order)
One main figure that is the finding. For most CPS papers this is a coefficient/forest plot, an
event-study plot around a reform, an RD plot, or a marginal-effects/predicted-probability plot — not
a wall of regression coefficients.
A clean main results table with the headline specification and the key robustness columns only;
push the full grid to the appendix.
A comparative descriptive exhibit showing the cross-case or over-time variation the design
exploits (so the reader sees the leverage).
A balance / first-stage / pre-trends exhibit appropriate to the design.
Make every exhibit self-contained
Title + caption state what is shown, the unit, the sample, and the comparison — readable alone.
Notes give: estimator, fixed effects, clustering level, N (units and observations), and what the
error bars are (CI level). For comparative panels, state the countries and the period.
Units and scales labelled in substantive terms (percentage points, predicted probability), not
raw coefficients a reader must decode.
Effect sizes visible. Show magnitudes and uncertainty, not just significance asterisks.
Comparative-specific exhibit craft
When comparing countries/regions, order and group them so the pattern is visible (by region,
regime type, or the key covariate) rather than alphabetically.
Use small multiples to show the same relationship across cases rather than one overplotted panel.
Maps only when geography is the point; otherwise a sorted dot plot communicates better.
Use when building tables and figures for a British Journal of Political Science (BJPS) manuscript so exhibits are self-contained, accessible, fit Cambridge's size limits, and earn their place under the word budget. Designs exhibits; it does not run the analysis.