DuckDB CLI skills

DuckDB CLI specialist for SQL analysis, data processing and file conversion. Use for SQL queries, CSV/Parquet/JSON analysis, database queries, or data conversion. Triggers on "duckdb", "sql", "query", "data analysis", "parquet", "convert data".

Security Scan

The skill is an instruction-only DuckDB CLI reference that asks for no credentials, installs, or unusual access and its requirements align with its stated purpose.

Purpose & Capability

Name/description promise SQL, data conversion and DuckDB CLI usage; the SKILL.md contains only DuckDB CLI commands, flags, and examples — nothing unrelated is requested.

Instruction Scope

Instructions direct the agent to run DuckDB CLI commands and to read/write local data files (CSV/Parquet/JSON) and ~/.duckdbrc; this is expected for a DuckDB CLI helper but means the agent will operate on any local files you point it at — review file paths and data provided to avoid exposing sensitive files.

Install Mechanism

No install spec or remote downloads are present (instruction-only skill), so nothing will be written or fetched by the skill itself.

Credentials

The skill declares no environment variables, credentials, or config paths beyond the normal ~/.duckdbrc reference; there is no disproportionate credential request.

Persistence & Privilege

always is false and the skill is user-invocable. It does not request persistent/system-wide changes or access to other skills' configs.

Assessment

This skill is an offline reference for using the DuckDB CLI and appears internally consistent. Before using: ensure DuckDB is installed on the host and on PATH, and only point the tool at files you intend to process (avoid running queries that read sensitive system files). Because the skill instructs the agent how to run shell commands, review any generated commands before execution and do not paste secrets or credentials into queries. If you want the agent to run commands autonomously, limit its scope or review outputs first.

DuckDB CLI Specialist

Helps with data analysis, SQL queries and file conversion via DuckDB CLI.

Quick Start

Read data files directly with SQL

# CSV
duckdb -c "SELECT * FROM 'data.csv' LIMIT 10"

# Parquet
duckdb -c "SELECT * FROM 'data.parquet'"

# Multiple files with glob
duckdb -c "SELECT * FROM read_parquet('logs/*.parquet')"

# JSON
duckdb -c "SELECT * FROM read_json_auto('data.json')"

Open persistent databases

# Create/open database
duckdb my_database.duckdb

# Read-only mode
duckdb -readonly existing.duckdb

Command Line Arguments

Output formats (as flags)

Flag	Format
-csv	Comma-separated
-json	JSON array
-table	ASCII table
-markdown	Markdown table
-html	HTML table
-line	One value per line

Execution arguments

Argument	Description
-c COMMAND	Run SQL and exit
-f FILENAME	Run script from file
-init FILE	Use alternative to ~/.duckdbrc
-readonly	Open in read-only mode
-echo	Show commands before execution
-bail	Stop on first error
-header / -noheader	Show/hide column headers
-nullvalue TEXT	Text for NULL values
-separator SEP	Column separator

Data Conversion

CSV to Parquet

duckdb -c "COPY (SELECT * FROM 'input.csv') TO 'output.parquet' (FORMAT PARQUET)"

Parquet to CSV

duckdb -c "COPY (SELECT * FROM 'input.parquet') TO 'output.csv' (HEADER, DELIMITER ',')"

JSON to Parquet

duckdb -c "COPY (SELECT * FROM read_json_auto('input.json')) TO 'output.parquet' (FORMAT PARQUET)"

Convert with filtering

duckdb -c "COPY (SELECT * FROM 'data.csv' WHERE amount > 1000) TO 'filtered.parquet' (FORMAT PARQUET)"

Dot Commands

Schema inspection

Command	Description
.tables [pattern]	Show tables (with LIKE pattern)
.schema [table]	Show CREATE statements
.databases	Show attached databases

Output control

Command	Description
.mode FORMAT	Change output format
.output file	Send output to file
.once file	Next output to file
.headers on/off	Show/hide column headers
.separator COL ROW	Set separators

Queries

Command	Description
.timer on/off	Show execution time
.echo on/off	Show commands before execution
.bail on/off	Stop on error
.read file.sql	Run SQL from file

Editing

Command	Description
.edit or \e	Open query in external editor
.help [pattern]	Show help

Output Formats (18 available)

Data export

• csv - Comma-separated for spreadsheets
• tabs - Tab-separated
• json - JSON array
• jsonlines - Newline-delimited JSON (streaming)

Readable formats

• duckbox (default) - Pretty ASCII with unicode box-drawing
• table - Simple ASCII table
• markdown - For documentation
• html - HTML table
• latex - For academic papers

Specialized

• insert TABLE - SQL INSERT statements
• column - Columns with adjustable width
• line - One value per line
• list - Pipe-separated
• trash - Discard output

Keyboard Shortcuts (macOS/Linux)

Navigation

Shortcut	Action
Home / End	Start/end of line
Ctrl+Left/Right	Jump word
Ctrl+A / Ctrl+E	Start/end of buffer

History

Shortcut	Action
Ctrl+P / Ctrl+N	Previous/next command
Ctrl+R	Search history
Alt+< / Alt+>	First/last in history

Editing

Shortcut	Action
Ctrl+W	Delete word backward
Alt+D	Delete word forward
Alt+U / Alt+L	Uppercase/lowercase word
Ctrl+K	Delete to end of line

Autocomplete

Shortcut	Action
Tab	Autocomplete / next suggestion
Shift+Tab	Previous suggestion
Esc+Esc	Undo autocomplete

Autocomplete

Context-aware autocomplete activated with Tab:

• Keywords - SQL commands
• Table names - Database objects
• Column names - Fields and functions
• File names - Path completion

Database Operations

Create table from file

CREATE TABLE sales AS SELECT * FROM 'sales_2024.csv';

Insert data

INSERT INTO sales SELECT * FROM 'sales_2025.csv';

Export table

COPY sales TO 'backup.parquet' (FORMAT PARQUET);

Analysis Examples

Quick statistics

SELECT
    COUNT(*) as count,
    AVG(amount) as average,
    SUM(amount) as total
FROM 'transactions.csv';

Grouping

SELECT
    category,
    COUNT(*) as count,
    SUM(amount) as total
FROM 'data.csv'
GROUP BY category
ORDER BY total DESC;

Join on files

SELECT a.*, b.name
FROM 'orders.csv' a
JOIN 'customers.parquet' b ON a.customer_id = b.id;

Describe data

DESCRIBE SELECT * FROM 'data.csv';

Pipe and stdin

# Read from stdin
cat data.csv | duckdb -c "SELECT * FROM read_csv('/dev/stdin')"

# Pipe to another command
duckdb -csv -c "SELECT * FROM 'data.parquet'" | head -20

# Write to stdout
duckdb -c "COPY (SELECT * FROM 'data.csv') TO '/dev/stdout' (FORMAT CSV)"

Configuration

Save common settings in ~/.duckdbrc:

.timer on
.mode duckbox
.maxrows 50
.highlight on

Syntax highlighting colors

.keyword green
.constant yellow
.comment brightblack
.error red

External Editor

Open complex queries in your editor:

.edit

Editor is chosen from: DUCKDB_EDITOR → EDITOR → VISUAL → vi

Safe Mode

Secure mode that restricts file access. When enabled:

• No external file access
• Disables .read, .output, .import, .sh etc.
• Cannot be disabled in the same session

Tips

• Use LIMIT on large files for quick preview
• Parquet is faster than CSV for repeated queries
• read_csv_auto and read_json_auto guess column types
• Arguments are processed in order (like SQLite CLI)
• WSL2 may show incorrect memory_limit values on some Ubuntu versions