Generate ad-hoc Python scripts with LLM and UV

The llm command-line tool combined with uv's inline dependency management enables rapid Python script creation without manual environment setup.

Setting Up the Template

First, install a script generation template for llm:

llm templates edit scripter

Add this template:

model: gpt-4.1
system: |
    You write Python tools as single file script.
    They always start with this comment:
    # /// script
    # requires-python = ">=3.12"
    # ///

    These files can include dependencies on python packages from PyPI.
    If they do, those dependencies are included in a list like this one in that same comment (here showing two dependencies):
    # /// script
    # requires-python = ">=3.12"
    # dependencies = [
    #     "click",
    #     "sqlite-utils",
    # ]
    # ///

    Don't add any text before of after the script.
    Don't quote the script in ``` or anything similar.
    Just output the Python code so that it can be saved directly into a .py file.

Generating and Running Scripts

Generate running llm with the -t parameter for using a template and piping the output into a .py file:

llm -t scripter "script that counts from 1 to 123 and indicates all prime numbers with a '*'" > primes.py

After reviewing the generated code, run it with:

uv run primes.py

UV automatically handles dependency installation based on the script's header comments — no virtual environment setup required.

Security Considerations

Always review generated scripts before execution. LLM-generated code may contain:

• Unintended behaviors or logic errors

• Security vulnerabilities

• Resource-intensive operations

• File system modifications

This workflow excels for quick prototypes and one-off utilities, but treat all generated code as untrusted until manually verified.