Use when analyzing the material of an Accounting, Organizations and Society (AOS) manuscript — coding and interpreting qualitative field data, estimating experimental and survey models, or running theory-laden archival analyses, with an audit trail appropriate to each tradition. Analyzes and reports; it does not design the study (aos-methods).
SkillScore86Maintained
Automatic
◆Research pack
Last verified 8/15/2026
What it changes
Passed 19 of 22 test cases with the skill, and 3 of 22 without it.
Without skill14%
With skill86%
Checks it passed that the model alone missed
The output includes the bracketed header 【Strand(s)】.
The output uses the bracketed header 【Process tests】.
The output uses the bracketed header 【Evidence chain】.
and 13 more in the full report
18 fixed1 passed either way1 still failing2 broke
Pass-rate lift
+73 pp
Tokens
+18%
Agent turns
0%
Response time
-28%
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 18 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 18
Prompt
We surveyed employees nested within 12 business units. Standard panel regression yields significant coefficients, but we worry about biased standard errors due to small cluster counts. What specialized correction technique should we implement in our script? Respond using the required bracketed output structure.
What was checked, without and with the skill
Checked
Without skill
With skill
The output includes the bracketed header 【Process tests】.
Missed
Passed
The response specifies executing a wild cluster bootstrap when cluster counts are small.
Passed
Passed
case-11 · 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.md82 linessingle file
Analysis & Evidence Craft (aos-data-analysis)
When to trigger
Interviews, observations, and documents are collected and must become findings
Experimental or survey data are in and the estimation plan is unsettled
A mixed-methods paper needs its qualitative and quantitative strands reconciled
Reviewers will probe the traceability of interpretations or the fit of the statistics
Code with the lens on. First-order codes stay close to informants' language; second-order categories translate them into the paper's theoretical vocabulary; the movement between the two is the analysis. Keep the codebook versioned so you can show how categories evolved.
Work the anomalies. The abductive engine: catalog episodes the current theory cannot absorb, and let them force conceptual revision — this is where AOS papers earn their contribution.
Build the evidentiary chain. Every conceptual claim should trace to identifiable material (interview number, meeting observed, document). Maintain a claim → evidence register; reviewers increasingly expect a data-structure or evidence table.
Weigh counter-evidence. Report material that resists the interpretation and say why the reading survives; interpretive rigor at AOS is demonstrated, not asserted.
Historical work: corroborate across independent archives; date claims precisely; distinguish what the sources show from what the genealogy argues.
Experimental and survey analysis
Match the model to the randomized design: ANOVA/ANCOVA with planned contrasts for factorial experiments; report cell means, standard deviations, per-cell n, and effect sizes, not p-values alone.
Test the theorized process: mediation with bootstrapped confidence intervals; moderation exactly as predicted, with simple-effects follow-ups.
Respect the randomization unit; report manipulation-check results and pre-specified exclusions transparently.
Surveys: reliability and validity evidence (alpha/CR, factor structure), common-method-bias diagnostics, and models matched to the nesting of the data.
Archival-with-theory analysis
Standard panel hygiene (fixed effects suited to the institutional claim, standard errors clustered to the data structure, documented sample screens) — but keep the estimand tied to the organizational/institutional construct, and interpret magnitudes in the theory's terms rather than as pricing effects.
Execution bridge (StatsPAI / Stata MCP)
For the quantitative lane of an AOS paper — experiments, surveys, and archival-with-theory
designs — execute and audit rather than only specify. Full map:
execution-with-mcp.
AOS is mixed-methods: route only the statistical strand through this bridge, and let the
qualitative strand keep its own audit trail (codebook, evidence register) outside it.
detect_design → recommend → fit with as_handle=true → audit_result to enumerate
the checks the design owes before a reviewer asks.
Experiments / many outcomes: randomization-based inference plus romano_wolf or
benjamini_hochberg for the multi-outcome families behavioral reviewers flag.
Use when designing the study for an Accounting, Organizations and Society (AOS) manuscript — qualitative field studies (access, case logic, interviews, observation), behavioral experiments, surveys, historical/archival inquiry, or theory-laden archival designs, each with the ethics and site-anonymity groundwork done early. Designs the inquiry; it does not run the coding or estimation (aos-data-analysis).