saineshnakra/automated-data-analyst

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "automated-data-analyst",
  "version": "0.5.0",
  "description": "Data analysis where every number comes from the ADA Python package (pandas, SciPy, statsmodels) with its calculation: a first look at any CSV or Excel table, and a paper's data audit, tables, figures and methods draft built from a study.yml and verified.",
  "author": {
    "name": "Sainesh Nakra",
    "url": "https://sainesh.com/"
  },
  "homepage": "https://github.com/saineshnakra/automated-data-analyst",
  "repository": "https://github.com/saineshnakra/automated-data-analyst",
  "license": "MIT",
  "keywords": [
    "data-analysis",
    "statistics",
    "research",
    "reproducible-research",
    "csv",
    "excel",
    "pandas",
    "scipy",
    "statsmodels",
    "mixed-models",
    "publication-figures"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "Automated Data Analyst",
        "shortDescription": "Every number computed in code",
        "longDescription": "Automated Data Analyst adds a skill that has the assistant analyze data with ADA, an open-source Python package, instead of doing the arithmetic itself. ADA computes every number with pandas, SciPy and statsmodels and prints the calculation behind it, or refuses with a reason when a number can't be computed honestly. The assistant reports ADA's numbers, each with its test or calculation, and passes refusals on.\n\nFor a quick look, ADA reads a CSV or Excel table and reports what is in it: data-quality issues with their rows, group comparisons, relationships and trends. For a paper, the assistant asks the researcher what only they know (what a row is, the subject column, the design, the planned analyses), writes a study.yml, and runs ada build, which makes a data audit, a data dictionary, descriptive and results tables in CSV, Word and LaTeX, figures with their source data and captions, and a methods draft. ada verify rebuilds everything from study.yml and the data and compares every number.\n\nThe skill needs a shell with Python 3.11 or newer where pip install automated-data-analyst works, as in Codex. ADA runs on your machine and its command line makes no network calls, so your data stays where it is.",
        "developerName": "Sainesh Nakra",
        "category": "Productivity",
        "capabilities": [
          "Report what is in a CSV or Excel table, each finding with its test and arithmetic",
          "Check data quality: missing values, duplicates and values outside their range, with their rows",
          "Compare groups with the test the study design calls for: t-tests, ANOVA or mixed models",
          "Write a study.yml from the data and the researcher's answers",
          "Build a paper's data audit, tables, figures and methods draft from study.yml",
          "Verify a build by rebuilding it and comparing every number",
          "Refuse with a reason when a number can't be computed honestly"
        ],
        "websiteURL": "https://github.com/saineshnakra/automated-data-analyst",
        "supportURL": "https://github.com/saineshnakra/automated-data-analyst/issues",
        "defaultPrompt": [
          "Use ADA to look at my data file and tell me what is in it, with the calculation behind each finding.",
          "Use ADA to set up a study from my data and build the paper's tables and figures.",
          "Use ADA to check my data for problems before I analyze it."
        ],
        "brandColor": "#635BFF",
        "composerIcon": "./assets/logo.png",
        "logo": "./assets/logo.png"
      }
    }
  }
}