Course Data Analysis Packager

Use this skill to turn a course topic and dataset into a complete, reproducible analysis deliverable. The expected result is a folder that another student, teacher, or teammate can understand, run, and present.

Status: experimental. For the first real use on a new course assignment type, produce a package plan first and ask for confirmation before creating many files, moving data, or rewriting an existing project.

Core Workflow

1. Clarify the assignment:

   • course, topic, deadline, required format, rubric, data source, group role, expected depth, and whether AI assistance is allowed.
   • if rules are missing, preserve uncertainty rather than inventing requirements.

2. Inspect data and sources:

   • identify raw files, schemas, missing values, duplicate rows, text fields, categorical fields, numerical fields, and privacy-sensitive columns.
   • for spreadsheets, use structured readers instead of ad hoc text parsing.

3. Define the research question:

   • make it narrow enough to answer with available data.
   • include comparison dimensions such as group, city, platform, time, user segment, topic, or risk level when useful.

4. Create the package structure:

   • read references/package-structure.md before creating or reorganizing files.
   • keep raw data separate from processed data.

5. Build the analysis plan:

   • choose metrics, charts, tables, assumptions, and limitations.
   • read references/analysis-workflow.md when writing scripts or notebook steps.

6. Produce deliverables:

   • README, data dictionary, analysis scripts/notebooks, output tables, figures, report template, and PPT outline.
   • keep figures directly usable in slides.

7. Run QA:

   • read references/deliverable-qa.md before final delivery.
   • verify commands run, outputs exist, claims match data, and privacy-sensitive fields are not exposed unnecessarily.

Output Modes

• package-plan: propose folder structure, research question, metrics, scripts, and deliverables.
• build-package: create or edit the actual folder package.
• review-package: inspect an existing package for missing files, weak claims, broken scripts, or presentation gaps.
• ppt-outline: turn analysis results into a slide-by-slide presentation outline.

Default Folder Contract

project_name/
  README.md
  data/
    raw/
    processed/
  docs/
    data_dictionary.md
    report_template.md
    ppt_slide_outline.md
  scripts/
  notebooks/
  output/
    tables/
    figures/

Use this as a default, but adapt to the existing project if it already has a coherent structure.

Writing Rules

• Use Chinese by default for Chinese course deliverables.
• Keep method names, data fields, and package names in English when that improves precision.
• Do not overclaim causal conclusions from descriptive or observational data.
• State missing data and sampling limits plainly.
• For platform or social-media data, include ethics and compliance notes when scraping, cookies, personal data, or platform rules are involved.
• Prefer simple, defensible analysis over complicated models that cannot be explained in class.

Final Response Contract

When delivering a completed package, report:

• folder path;
• key files created or updated;
• commands to reproduce;
• main charts/tables produced;
• remaining assumptions or limitations;
• what still needs manual verification.