Ingest S3 Data Safely with Bauplan Skills for Claude Code

Import S3 data into Bauplan with Claude Code, using an isolated feature branch so nothing lands in main until it has been checked.

This walkthrough starts from an empty uv project and Claude Code with the Bauplan Skills plugin installed. From a single prompt, Claude reads a Parquet file from S3, asks how the table should be named and when validation should happen, writes a reusable ingestion script, and runs it. 109,000 rows land on a feature branch, not on main.

From there we inspect the data on the branch, review the 13-column schema, merge the branch into main, delete the branch, and run a SELECT COUNT to confirm the merge. The same table is then verified visually in the Bauplan UI, with metadata and a data preview.The second half covers appending. A new prompt points Claude at a second Parquet file in S3. Claude finds the ingestion script it wrote earlier, uses it as the reference to generate an append script, updates pyproject.toml, runs it, and we query the table again to confirm the append.

What you will see:

• Installing the Bauplan Skills plugin via the Anthropic marketplace and confirming it in Claude Code
• Prompting an S3 to Bauplan import with table naming and validation choices
• Data landing on an isolated feature branch instead of main
• Inspecting branch data and reviewing the schema before merge
• Merging to main, deleting the feature branch, and validating with a query
• Reviewing the generated ingestion script as a durable, reusable artifact
• Visual validation of the table in the Bauplan UI
• Appending a second Parquet file by reusing the earlier artifact