In AI-assisted development, repository structure is part of the context agents need to work reliably. Project-level instructions and source access can make an agent competent within one project, yet leave it without the system context needed across services and teams. section 10.0
Repository structure creates practical tradeoffs for agent navigation:
- A monorepo can expose more code, but visibility alone does not make the code easy to navigate. section 10.0
- Without topology and dependency information, agents may find files but miss how they connect to the larger system. section 10.2
- Wrong turns consume tokens and reviewer attention; repeated navigation failures turn infrastructure gaps into team costs. section 10.2
The wrong checkout package
An agent infers that checkout-v2-experiment is the active implementation from commit dates, overlooking stable production code in packages/checkout. The mistake is an information failure: the repository gives the agent no clear way to distinguish active code from abandoned experiments. section 10.2













