John Locke Describes Multi-Repository Workflow for AI-Assisted Development

Managing AI coding assistants across multiple repositories becomes difficult when a feature touches deployment pipelines, monitoring systems, automation tools, and application code. In a blog post published on 2 June 2026 on Freelock, founder and lead developer John Locke described Argo, a workflow that groups related repositories beneath a shared parent project. The structure is designed to give coding agents broader project context while preserving repository-level guidance.

According to Locke, Argo places multiple internal repositories under a common parent directory containing shared documentation, architectural guidance, development principles, and agent instructions. A root CLAUDE.md file provides project-wide context, while child repositories can maintain their own instruction files for local conventions, commands, and implementation patterns. Locke also uses an AGENTS.md file as a symlink to CLAUDE.md for Opencode, repeating the pattern across child repositories.

The post also describes context markers, a pattern borrowed from Lada Kesseler’s Augmented Coding Patterns, to show whether an agent has loaded the relevant instruction files. The root Argo file uses one marker, while sub-repositories such as carina/, pipelines/, and watney/ add their own markers to indicate repository-specific context. Locke frames the approach as machine-readable project documentation that helps expose context boundaries, though the post does not provide quantitative measures of its effect on development outcomes.