ai4ss-skills

Develop and revise working academic prose for empirical political-science research. Use when clarifying a paper's central point, building or restructuring an argument, drafting a working introduction, theory, research-design, results, or conclusion section, integrating literature and evidence, revising an author draft, or diagnosing why a paper feels unfocused. Triggers include "academic writing", "working paper", "draft this section", "argument outline", "revise the paper", "论文写作", "论文初稿", "论证结构", "修改论文".

Turn completed quantitative analyses into rigorous political-science interpretation for researchers and collaborators. Use when a project has multiple tables, figures, models, diagnostics, or null and conflicting results that need to be understood together; when effect magnitudes need substantive interpretation; or when the team must decide what the evidence changes and what to do next. Triggers include "explain the results", "interpret the analysis", "summary for coauthors", "what do these findings mean", "解释结果", "结果解读", "给合作者的分析总结".

Second step of the DDI survey-cleaning harness. Reads `ddi-metadata.yaml` (produced by `codebook-parse`) and guides the researcher in declaring all recoding decisions as DDI Lifecycle 3.3–compliant metadata in a new `cleaning_contract` block — before any data is touched. Trigger on any of: "declare my recoding decisions", "write the cleaning contract", "set up recodes", "which variables need recoding", "how should I handle missing codes", "set up my analysis variables", "I want to recode this survey", "configure the cleaning step", "define universe filter", "set missing values", or any time the user has `ddi-metadata.yaml` and wants to specify how to clean it before running `cleaning-execute`. Precondition: `ddi-metadata.yaml` must exist in the working directory. If it does not, tell the user to run `codebook-parse` first.

Third step of the DDI survey-cleaning harness. Reads `ddi-metadata.yaml` (with a `cleaning_contract` block written by `cleaning-contract`) and executes it mechanically: produces a clean `<stem>-clean.csv`, a reproducible `<stem>-cleaning.R` script, and appends a full audit trail to `processing_events`. Trigger on any of: "execute the cleaning contract", "run the cleaning", "produce the clean dataset", "generate the cleaned CSV", "apply my recoding decisions", "execute cleaning_contract", "build the analysis dataset", "run cleaning-execute", or any time the user has a `ddi-metadata.yaml` with a `cleaning_contract` block and wants the cleaned data + R script. Precondition: `ddi-metadata.yaml` must exist AND contain a non-empty `cleaning_contract` block. If it does not, tell the user to run `cleaning-contract` first.

First step of the DDI survey-cleaning harness. Reads a survey dataset or codebook file and produces `ddi-metadata.yaml` — a DDI Lifecycle 3.3–compliant Single Source of Truth (SSOT) that all downstream skills consume. Trigger on any of: "parse this dataset", "generate codebook metadata", "extract variable labels", "create the DDI YAML", "start the cleaning workflow", "document this survey file", "what variables does this dataset have", "parse this questionnaire PDF", "read the codebook Word file", or any time a user provides a survey file and wants to begin cleaning, labelling, or reproducible analysis. Supported inputs: Stata .dta, SPSS .sav, SAS .sas7bdat, CFPS YAML wrapper, DDI Codebook 2.5 XML, CSV with a data dictionary sidecar, PDF questionnaire or codebook, Word .docx variable documentation.