A dependency is a component or package your software relies on. Reusing a tested library can save time and avoid defects, but an install is not just a shortcut to functionality. It adds a versioning surface, a security surface, a behavioral contract outside your control, and possibly a tree of transitive packages. The decision is whether the reuse remains worthwhile over the system’s lifetime. section 4.3
Evaluate a proposed dependency through three questions: Will users get better functionality, performance, or reliability? Will faster delivery still pay off after patches and upgrades? Will the package make the system easier to maintain, or merely move code out of the repository? Foundational dependencies may justify substantial upkeep; a small convenience package may not. section 4.3.1
One function, a larger surface
Suppose a requirement calls for one email check. Importing a validation library is quick, but it may bring unused validators and transitive packages into the build. A small, auditable function may fit the stated requirement instead. Neither choice is automatic: check what the function must accept, what it must reject, and how those requirements might change. Packages you never import directly can still require security patches or break a build. section 4.3.2 section 4.4.1
AI can create a second kind of dependency: reliance on generated code and decisions that developers may not understand well enough to debug. It can also introduce external packages without accounting for their maintenance cost. In one styling example, an AI-suggested package establishes a competing pattern alongside existing CSS modules; later suggestions repeat it, leaving two styling systems to support. section 4.3.3 section 4.3.4
Before accepting a dependency:
- State the required behavior and identify packages or patterns that need justification rather than adding them by default. section 4.4.1
- Compare reuse with a small local implementation; weigh delivery time against transitive packages, security work, upgrades, and debugging. section 4.3.1 section 4.3.2
- Keep chosen dependencies isolated and explicit, so they are easier to disable or remove; review AI suggestions against existing standards. section 4.6 section 4.3.4


































