R analysis

Skill alexvantwisk/supeRpowers/skills/r-analysis

Guided data analysis pipeline — import, clean, explore, model, visualize, reportFrom its SKILL.md

Install

npx -y skills add alexvantwisk/supeRpowers --skill r-analysis

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.1 KB, as published. Nobody here has run it

Analysis Pipeline

Guided workflow: import, clean, explore, model, visualize, report.

Prerequisites

Data file accessible (CSV, Excel, database, or API)
Clear research question or analysis goal
R session with tidyverse installed

Progress Tracking

Use TaskCreate at the start of this workflow — one task per phase below. Mark each in_progress when starting, completed when its gate passes.

"Phase 1: Import data"
"Phase 2: Clean data"
"Phase 3: Explore data"
"Phase 4: Model"
"Phase 5: Visualize findings"
"Phase 6: Report"

Surfaces the orchestration shape to the user and prevents skipped phases or premature completion claims.

Steps

Step 1: Import

Skill: r-data-analysis Action: Read data with explicit col_types. Inspect with glimpse() and skimr::skim(). Document dimensions, column types, and any import warnings. Gate: Data loaded successfully. Row count and column types confirmed.

Step 2: Clean

Skill: r-data-analysis Action: Handle missing values (diagnose with across(everything(), \(x) sum(is.na(x)))). Fix types, rename columns to snake_case, remove duplicates. Document cleaning decisions. Gate: No unexpected NAs in key columns. Types correct. No duplicates in ID columns.

Step 3: Explore

Skill: r-data-analysis + r-visualization Action: Compute summary statistics (.by grouping). Plot distributions, relationships, and potential outliers. Identify key patterns and form hypotheses. Gate: Key patterns identified. Research questions refined.

Step 4: Model

Skill: r-stats Agent: r-statistician — dispatch for model selection, assumption verification, or interpretation guidance Action: Fit appropriate model(s). Check assumptions (residual plots, normality, homoscedasticity). Interpret coefficients and confidence intervals. Gate: Model assumptions met (or documented violations with justification). Results interpretable.

Step 5: Visualize

Skill: r-visualization Action: Create publication-quality plots of key findings using ggplot2. Use patchwork for multi-panel layouts. Apply consistent theme. Gate: Plots communicate findings clearly. Axes labeled, legends readable.

Step 6: Report

Action: Summarize findings in plain language. Document limitations, assumptions, and next steps. Structure as: question → method → result → interpretation. Gate: Narrative matches evidence. Limitations acknowledged.

Abort Conditions

Data quality too poor to proceed (>50% missing in key variables) — report to user, suggest data collection improvements.
Model assumptions severely violated with no reasonable alternative — document and present descriptive analysis only.
Statistical findings contradicted by exploratory analysis — re-examine data and model, do not proceed until resolved.

Examples

Example: Exploratory analysis of a CSV dataset

Prompt: "Analyze sales_data.csv — what drives revenue by region?"

Flow: Import (read_csv with col_types) → Clean (handle NAs in revenue, fix date types) → Explore (revenue distributions by region, correlation with predictors) → Model (linear regression: revenue ~ region + season + marketing_spend) → Visualize (coefficient plot + regional trends) → Report

# Step 1 — Import
sales <- readr::read_csv("sales_data.csv", col_types = cols(
  region = col_character(), date = col_date(),
  revenue = col_double(), marketing_spend = col_double()
))

# Step 3 — Explore
sales |>
  summarise(
    mean_revenue = mean(revenue, na.rm = TRUE),
    n = n(),
    .by = region
  )

Example: Quick EDA on a new dataset

Prompt: "I just got this dataset, help me understand what's in it."

Flow: Import → Clean (light touch) → Explore (heavy — distributions, missing patterns, relationships) → Skip Model → Visualize (overview plots) → Report (summary of what the data contains and potential questions)