Repeatability
High
This is a well-understood software engineering pattern — wrapping a data-fetching library with configuration-driven TTL logic is structurally the same problem every time. The agent can follow established patterns for SWR/TanStack Query customization without novel judgment.
Ambiguity Tolerance
Medium
The core requirements (per-endpoint TTLs, stale-while-revalidate, background refetch, clean component API) are crisp enough to code against. However, specifics like the exact config schema, TypeScript generics depth, and error handling strategy leave room for interpretation that may not match team expectations.
Data & Tool Availability
Medium
The agent needs access to the existing codebase, the list of 8 endpoints with their TTL requirements, and the current data-fetching setup. If these are provided, the task is executable; if the agent must infer them from partial context, quality degrades.
Error Cost
Low
This is new wrapper code, not a modification to production data or live systems. Mistakes produce incorrect caching behavior that is observable in testing and easily corrected before deployment. The reversibility is high.
Human Judgment Required
Low
The task is technical and pattern-driven with no ethical, relational, or taste-based dimensions. A human review pass for code style and architectural fit is advisable but not strictly required for a functional output.