An Owned Command Matrix for Setup Guides That Models Must Not Extend

The process of creating setup guides for software installations involves a structured approach to ensure the documentation remains accurate and reliable. This process involves two primary lanes: a matrix lane for storing verified facts and a narrative lane for crafting the explanatory content. The matrix lane contains a synthetic JSON file that lists supported installation paths, each with specific details such as the operating system, package manager, minimum version, installation command, expected exit code, verification date, fixture output file, and the human owner responsible for maintaining the row.

The narrative lane, on the other hand, is where model drafts can create sections, explain the purpose of each step, and propose troubleshooting hypotheses, all of which must cite the matrix values rather than invent new information. The key rule is that any conflict between the narrative and the matrix is resolved in favor of the matrix, ensuring that all claims are grounded in verified facts.

Ownership of critical operational details like the operating system label, package manager, and exact installation command lies solely with designated humans, who are responsible for updating these entries as needed. The workflow involves freezing rows for each supported installation path before starting the drafting process, capturing the actual fixture output from running the commands without any cleanup by models, and rejecting any unknown shell commands to maintain the integrity of the documentation.

This structured approach helps prevent the creation of misleading or incomplete installation guides by ensuring that all claims are traceable back to verified installation outputs.