convert-file

---

name: convert-file description: > Convert any data file to another format: CSV, Parquet, JSON, Excel, GeoJSON, and more. Use when the user says "convert to parquet", "save as xlsx", "export as JSON", "make this a CSV", "turn into parquet", or any variation of format-to-format conversion for data files. Also triggers when the user wants to write Parquet, Excel, or other binary formats that Claude cannot produce natively. argument-hint: [output-file] allowed-tools: Bash

You are helping the user convert a data file from one format to another using DuckDB.

Input file: $0 Output file: ${1:-}

Step 1 — Resolve input and output

Input: $0. If it's a bare filename (no /), resolve to a full path with find "$PWD" -name "$0" -not -path '*/.git/*' 2>/dev/null | head -1.

Output: If $1 is provided, use it as the output path. If not, default to the same stem as the input with a .parquet extension (e.g., data.csv → data.parquet).

Infer the output format from the output file extension:

Extension	Format clause
.parquet, .pq	(default, no clause needed)
.csv	(FORMAT csv, HEADER)
.tsv	(FORMAT csv, HEADER, DELIMITER '\t')
.json	(FORMAT json, ARRAY true)
.jsonl, .ndjson	(FORMAT json, ARRAY false)
.xlsx	(FORMAT xlsx) — requires INSTALL excel; LOAD excel;
.geojson	(FORMAT GDAL, DRIVER 'GeoJSON') — requires LOAD spatial;
.gpkg	(FORMAT GDAL, DRIVER 'GPKG') — requires LOAD spatial;
.shp	(FORMAT GDAL, DRIVER 'ESRI Shapefile') — requires LOAD spatial;

Step 2 — Convert

Run a single DuckDB command. Prepend extension loads as needed based on both the input and output formats.

duckdb -c "
<EXTENSION_LOADS>
COPY (FROM '<INPUT_PATH>') TO '<OUTPUT_PATH>' <FORMAT_CLAUSE>;
"

For remote inputs (s3://, https://, etc.), prepend the same protocol setup as read-file:

Protocol	Prepend
s3://	LOAD httpfs; CREATE SECRET (TYPE S3, PROVIDER credential_chain);
gs:// / gcs://	LOAD httpfs; CREATE SECRET (TYPE GCS, PROVIDER credential_chain);
https:// / http://	LOAD httpfs;

If the user mentions partitioning (e.g., "partition by year"), add PARTITION_BY (col) to the format clause. This only works with Parquet and CSV output.

If the user mentions compression (e.g., "use zstd"), add CODEC 'zstd' for Parquet output.

Step 3 — Report

On success, report:

• Input file and detected format
• Output file, format, and size (ls -lh)
• Row count if quick to compute

On failure:

• duckdb: command not found → delegate to /duckdb-skills:install-duckdb
• Missing extension → install it and retry
• Input parse error → suggest the user check the input format or try /duckdb-skills:read-file first to inspect it