Log-build

Add minimal, systematic, high-value logs that help humans and agents locate failures quickly.
Goal: maximize debugging value per log line, not log count.

Core policy

1. Detect first — check if the touched path actually needs logs.
2. Reuse existing infrastructure — logger, format, directory, level conventions.
3. If no logger exists and the flow is non-trivial — introduce the smallest shared setup (file-backed, timestamped, consistent format) instead of scattering ad hoc prints.
4. Add few but decisive logs at key nodes only.
5. If the task is trivial with no meaningful log point — add nothing.

Where to log