In child-directed speech corpora, a practical protocol is a two-pass process where a primary annotator completes all tiers and a verifier reviews only items auto flagged by rules such as low alignment confidence, cross-tier mismatches, or long gaps. This triage preserves phonetic and gesture detail because stable segments are locked, while flagged tokens get side-by-side inspection and notes. One rubric item that can improve agreement is a clear rule for gesture boundaries: onset is the first visible move away from rest toward the stroke, and offset is the return to rest or a sustained hold that ends the stroke. The template that supports this is a short checklist with forced-choice fields for gesture type (deictic, iconic, beat), anchor syllable, and certainty level, plus a note box that is only required when "uncertain" is chosen. Constraining the verifier's workload and standardizing boundary calls tend to raise kappa or alpha while cutting review time.