Use when the headline result of an American Economic Journal: Macroeconomics (AEJ: Macro) manuscript must be shown stable across specification, sample, identification, and tuning choices. Builds the robustness program a macro referee will demand; it does not establish the primary identification or model (use aejmac-identification / aejmac-theory-model first).
SkillScore86Maintained
Automatic
◆Research pack
Last verified 8/15/2026
What it changes
Passed 19 of 22 test cases with the skill, and 16 of 22 without it.
Without skill73%
With skill86%
Checks it passed that the model alone missed
The response specifies aejmac-tables-figures under Next step.
The response outputs aejmac-tables-figures in the Next step section.
The response uses full-width Japanese brackets 【】 for section headers.
and 6 more in the full report
5 fixed14 passed either way1 still failing2 broke
Pass-rate lift
+14 pp
Tokens
+20%
Agent turns
0%
Response time
-31%
Model gemini-3.6-flashMeasured 8/15/2026v119 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
After finalizing our macro robustness check plan and running all sensitivity tests, we need to format our final empirical tables and impulse response figures for submission. What is the designated next step skill handle to call?
What was checked, without and with the skill
Checked
Without skill
With skill
The response references aejmac-tables-figures as the designated next step handle.
Missed
Passed
case-20 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md88 linessingle file
Robustness Program (aejmac-robustness)
When to trigger
The headline number rests on one specification, one sample, one lag length, or one grid
A referee could ask "is this an artifact of [choice]?" and you have no panel of alternatives
The empirical IRF and the model-implied response are compared but only at the baseline
A structural/calibrated result has never been re-run under alternative targets
The AEJ: Macro robustness bar
Macro inference is fragile in characteristic ways: short effective samples, structural breaks (Great Moderation, ZLB, COVID), specification forks (lag length, detrending, prior, calibration target), and method dependence (SVAR vs. LP; perturbation vs. global). The AEJ: Macro robustness bar is to show the headline quantity survives the choices a skeptical macro referee would flip, and to be honest where it does not. Robustness is not a graveyard of extra tables — it is a targeted defense of the specific number the paper claims.
A macro robustness program (build the panel)
Empirical (SVAR / LP / narrative)
Sample splits: pre/post-1984 (Great Moderation), exclude/keep the ZLB period, exclude COVID; report whether the response is stable.
Specification: lag length, detrending/filtering choice (HP vs. one-sided vs. none), control set, levels vs. differences.
Method cross-check: if SVAR is baseline, corroborate with LP (and vice versa); agreement is strong evidence.
Inference: alternative HAC bandwidths / clustering; weak-instrument-robust bands for proxy-VAR/LP-IV.
Identification variants: alternative orderings / sign sets / instrument constructions.
Quantitative (DSGE / HANK / structural)
Alternative calibration targets and parameter ranges; show how the headline quantity moves.
Alternative solution method / accuracy (higher perturbation order, finer grid) where nonlinearity matters.
Alternative model elements (Taylor-rule coefficients, adjustment costs, market structure) the referee will name.
Estimation: alternative moments / priors; re-estimate on a subsample.
Cross-cutting
External validity: another country / dataset / period where the mechanism should also hold.
Placebo / falsification: a response that should be zero (pre-shock leads; a non-targeted series).
Reporting discipline
Lead with a one-paragraph summary of what is robust and what is not, then a compact robustness table/figure.
Keep the baseline number visible in every robustness exhibit so the reader sees the movement.
Put the bulk in the online appendix; main text carries the decisive checks only.
A spec-curve / multiverse plot is powerful for empirical macro when many forks exist.
Use when extensions, edge cases, or applied robustness checks are missing for an American Economic Journal: Microeconomics (AEJ: Micro) manuscript — covering theory extensions (relaxed assumptions, alternative concepts, perturbations) and applied/experimental robustness. Decides which extensions earn their place; it does not prove the main result (see aejmic-theory-model).