Where exactly does director intent end and cinematographer execution begin in a generation prompt?

In the founder's system the Director authors shotMood, keyLighting and performance while the DP edits scene/action/camera/emotion layers for the render. Which fields must stay intent-only (never specify a fixture or focal length) and which must be execution-only, so a DP agent retains technical and creative judgment? Why this is still open: Both roles use the word 'lighting' for their part; a field-level boundary is not written down (see roles.json C2).