data-wrangler

Data Wrangler Plugin for Claude Code

A Claude Code plugin that enables powerful data transformation and export using DuckDB SQL.

Features

• Multiple file formats: CSV, Parquet, JSON/NDJSON, Excel (.xlsx)
• Database connections: Postgres, MySQL, SQLite, S3
• Cloud storage: AWS S3, Azure Blob, Google Cloud Storage, Cloudflare R2
• Secrets management: Secure credential storage with environment variable substitution
• Direct file queries: Query files directly by path
• Aliased sources: Join multiple datasets with clean SQL
• Write mode: Export results to Parquet, CSV, or JSON
• Partitioning: Create Hive-style partitioned datasets
• Advanced SQL: PIVOT/UNPIVOT, sampling, window functions

Installation

From GitHub

Add the marketplace and install:

/plugin marketplace add richard-gyiko/data-wrangler-plugin
/plugin install data-wrangler@data-wrangler-marketplace

Local Development

/plugin marketplace add ./path/to/data-wrangler-plugin
/plugin install data-wrangler@data-wrangler-marketplace

Usage

Once installed, Claude will automatically use this skill when you ask data transformation questions.

Examples

Simple file query:

"What are the top 10 products by revenue in sales.csv?"

Join multiple files:

"Join orders.parquet with customers.csv and show total orders per customer"

Export to Parquet:

"Filter active users from users.csv and save as users_active.parquet"

Create partitioned dataset:

"Convert events.json to Parquet partitioned by year and month"

Secrets Management

Store database and cloud credentials securely using a YAML secrets file with environment variable substitution.

Create secrets.yaml:

secrets:
  my_postgres:
    type: postgres
    host: db.example.com
    user: analyst
    password: "${PGPASSWORD}"  # Uses environment variable
    database: analytics

Supported secret types:

• Databases: PostgreSQL, MySQL
• Cloud storage: AWS S3, Azure Blob, Google Cloud Storage, Cloudflare R2
• HTTP: Bearer token authentication
• Other: HuggingFace, Iceberg, DuckLake

See SECRETS.md for full configuration options.

How It Works

The skill uses DuckDB, an embedded analytical database that excels at:

• Columnar storage and vectorized execution
• Direct querying of files without loading into memory
• Automatic format detection
• Efficient aggregations on large datasets

Requirements

• Python 3.11+
• uv package manager (for running the script)

Dependencies are automatically installed via the inline script metadata:

• duckdb>=1.4.3
• polars[pyarrow]>=1.36.1

File Structure

data-wrangler-plugin/
├── .claude-plugin/
│   ├── plugin.json          # Plugin manifest
│   └── marketplace.json     # Marketplace definition
├── skills/
│   └── data-wrangler/
│       ├── SKILL.md         # Skill instructions for Claude
│       ├── SECRETS.md       # Secrets management docs
│       ├── TRANSFORMS.md    # Advanced SQL patterns
│       └── scripts/
│           └── query_duckdb.py  # DuckDB query engine
├── tests/                   # Test suite
└── README.md

License

MIT License - see LICENSE file for details.

Contributing

Contributions welcome! Please open an issue or pull request.