Using LLM in the shebang line of a script

2026-05-11 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

Simon Willison investigated methods for integrating the LLM CLI tool directly into shebang lines, inspired by a Hacker News comment suggesting English text files could use shebangs. The simplest approach leverages LLM fragments, allowing direct generation requests like creating an SVG of a pelican. More advanced patterns incorporate tool calls using the `-T name_of_tool` option, enabling LLM to execute specific functions, such as fetching the current time for a haiku. The article also demonstrates executing YAML templates directly via `llm -t`, which can define custom Python functions for calculations. An example shows a `calc.sh` script performing arithmetic operations by calling `add` and `multiply` functions defined within the YAML template, outputting the result along with tool debug information.

Key takeaway

For AI Engineers developing command-line utilities, consider embedding LLM CLI commands directly into shebang lines. This approach streamlines script execution for generative tasks or tool-augmented operations, reducing boilerplate. You can define custom Python functions within YAML templates for complex logic, making your scripts more powerful and self-contained. Explore the `-f`, `-T`, and `-t` options to integrate LLM capabilities efficiently into your shell scripts.

Key insights

LLM CLI can execute generative AI tasks and tool calls directly from a script's shebang line.

Principles

• Shebangs can execute LLM commands.
• LLM fragments simplify direct text generation.
• YAML templates enable custom tool definitions.

Method

Integrate `llm` with `-f` for fragments, `-T` for tools, or `-t` for YAML templates directly into a shebang line to execute generative AI tasks or custom functions.

In practice

• Generate SVGs directly from a script.
• Incorporate current time into generated text.
• Define custom arithmetic functions for scripts.

Topics

• LLM CLI tool
• Shebang line
• LLM fragments
• Tool calls
• YAML templates

Best for: AI Engineer, Software Engineer

Related on AIssential

• llm 0.32a0 — The `llm` CLI tool provides direct command-line access to large language models.
• Get Vision LLMs to Follow Your Rules: Prompt-Guided JSON Formatting — Prompt-guided JSON formatting enables LLMs to generate structured data with specific, user-defined output formats.
• Run a Local LLM with OpenClaw on Your Mac Mini — Running a local, quantized LLM on a Mac Mini with llama.cpp significantly reduces OpenClaw API costs while maintaining performance for common agent tasks.
• Ollama Python library 0.4 with function calling improvements — Ollama's 0.4 Python library simplifies tool integration by directly accepting Python functions and auto-generating JSON schemas.
• FasterPy: An LLM-based Code Execution Efficiency Optimization Framework — FasterPy optimizes Python code execution efficiency by combining LLMs with RAG and LoRA, outperforming existing methods.
• Smart Document Extraction with Business Rules — Gemma vs Qwen vs Ministral — LM hints in Sparrow enable LLMs to generate and format data fields dynamically during query execution.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.