Line-1 Deliverable Protocol (No-Preamble Code Blocks)

Overview

When asked to write a script or generate code ("Write a Python script to parse this CSV"), default LLMs output 1 to 3 lines of introductory "throat-clearing" before the code block ("Sure thing! I would be glad to help you with that. Here is the complete Python script that will parse your CSV file:").

Introductory preambles cause two major engineering issues:

1. Breaks CLI Unix Pipelines: Developers and scripts cannot pipe LLM outputs directly into runtime interpreters (llm "write script" | python) because conversational English text at Line 1 causes a SyntaxError: invalid syntax.
2. Output Token Waste: Emitting 25 to 50 tokens of polite preamble across dozens of queries wastes thousands of output tokens per week.

The Line-1 Deliverable Protocol mandates that Byte 0 of the model's output MUST be the opening code block backtick (```) or raw code stream.

Introductory Preamble vs. Line-1 Deliverable