Use when running and reporting the analysis for an Annals of the American Association of Geographers manuscript — spatial statistics and modeling, remote-sensing accuracy, or qualitative coding and interpretation. Sets analysis and reporting norms across the four areas; it does not choose the design.
SkillScore83Maintained
Automatic
◆Data Analyst pack
Last verified 8/15/2026
What it changes
Passed 19 of 23 test cases with the skill, and 15 of 23 without it.
Without skill65%
With skill83%
Checks it passed that the model alone missed
The output sets 【Mode】 to qualitative.
The output includes 【Headline result】.
The Next field points to aaag-tables-figures.
and 10 more in the full report
6 fixed13 passed either way2 still failing2 broke
Pass-rate lift
+17 pp
Tokens
+27%
Agent turns
0%
Response time
-20%
Model gemini-3.6-flashMeasured 8/15/2026v119 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
Audit my historical geography archive coding method using the designated structured output block format.
What was checked, without and with the skill
Checked
Without skill
With skill
The output sets 【Mode】 to qualitative.
Missed
Passed
The output includes 【Headline result】.
Missed
Passed
The output includes 【Reproducibility】 with a [Y/N] indicator.
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.md92 linessingle file
Data Analysis (aaag-data-analysis)
The Annals expects analyses that are spatially honest and reported with uncertainty, whatever the
area. The standard is that a competent reader in the area could follow the logic from data to claim and
see that the geography of the data was respected, not flattened.
When to trigger
Estimating models, running spatial statistics, classifying imagery, or coding qualitative material
A reviewer questioned uncertainty, robustness, spatial autocorrelation, accuracy, or interpretation
Preparing the results section and deciding what to report
Spatial / quantitative
Diagnose space first. Report spatial autocorrelation in residuals; if present, move to a spatial
model (lag/error, GWR/MGWR, spatial regimes) and say why.
Uncertainty everywhere. CIs/SEs (spatially robust where needed), not stars alone; for prediction,
out-of-sample error from spatial/blocked CV.
Robustness. Re-estimate across plausible areal units and bandwidths (MAUP/scale sensitivity);
show the result is not a unit artifact. Report effect sizes in interpretable units.
Remote sensing / physical
Accuracy with an independent sample. Confusion matrix, overall/producer/user accuracy, kappa or
F1; for continuous outputs, RMSE/MAE and bias; map the spatial pattern of error, not just a scalar.
Propagate uncertainty from inputs through to the reported quantity; state the validation design.
Qualitative / interpretive
Transparent analytic trail. Coding scheme, how themes were derived, and how interpretations were
checked (negative cases, member checks, triangulation) — credibility over counting.
Evidence-to-claim mapping. Each interpretive claim is tied to identifiable (anonymized) evidence;
avoid quote-mining that over-generalizes from one informant.
Mixed methods
Show the integration. State where the strands converge and where they conflict, and how the
conflict was adjudicated — do not report two parallel analyses and call it mixed methods.
Cross-cutting reporting bar
Match every claim in the text to an exhibit or statistic; no orphan assertions.
Use when handling data documentation, sharing, and spatial-data ethics for an Annals of the American Association of Geographers manuscript — provenance (sources, projections, processing), reproducibility, and geoprivacy / human-subjects protection. Prepares documentation; it does not over-state a verification gate the policy does not impose.