Rehearsal Plan Contract
Load for rehearsal, not slide generation. The SKILL owns initial pacing defaults; use measured speaker/venue timing to evaluate the rehearsal. Narrative and visual design remain separate recipes; Cue owns recording execution.
Plan and Measurement
For a 20–30 minute talk, the local planning target is 7 solo + 2 peer-audience + 1 venue pass; it is not an empirical guarantee. Record any venue/speaker-specific adjustment. Use early passes for content, later passes for timing, a recorded pass for measurement, peer passes for questions, and a final check on the actual equipment (target: 60 minutes before showtime).
Measure WPM, fillers/minute, pause count and section duration from a real recording. Completion targets: WPM within ±5 of the chosen target and fillers <1/minute. With no recording, report these as unverified goals, not achieved results. Mark pauses in speaker notes (defaults: 2 seconds after key claims, 3 seconds between sections). Budget pauses, demos, transitions and Q&A inside the hard clock.
Select venue-appropriate eye routing: lighthouse_5_zone (front-left/right, center, back-left/right; planned rotation), camera_only, or hybrid_30s_alt. These are plan choices, not proof of audience attention. Rehearse recovery without restarting from the beginning whenever a line is missed.
Q&A and Contingencies
Prepare known hard questions, an honest unknown-answer response and a backup self-question. Restate unclear questions; keep multipart/hostile exchanges within the Q&A budget rather than inventing an answer.
Document at least three fallback plans: slides unavailable (offline copy/no-slide summary), demo failure (recorded fallback), and reduced time (pre-marked cuts retaining the governing idea and CTA). Check microphone, clicker, display, sound and presenter view on actual equipment. Include recover-from-blank steps and a pre-talk routine suited to the speaker. Do not prescribe dietary/breath-holding rules or guaranteed physiological effects as slide-delivery requirements.
Output Schema
Placeholders are targets or measurements as labeled, never sample observations:
rehearsal_plan:
talk_duration_min: 0
target_wpm: 0
passes_solo: 7
passes_live: 2
passes_stage: 1
recordings:
- pass: 0
wpm: 0
filler_per_min: 0
pause_count: 0
pause_markers: []
eye_routing: "lighthouse_5_zone | camera_only | hybrid_30s_alt"
qa_prep:
self_qa_backup: "<prepared question>"
known_hard_questions: []
contingencies:
slides_fail: "<tested fallback>"
demo_fails: "<artifact/path and playback check>"
time_cut: "<explicit cuts and retained ending>"
pre_talk_routine: []Use recordings: [] until a recording exists; populate measurements only from evidence. Accompany the plan with measured-vs-target status, explicit pause locations, Q&A backup, ≥3 contingencies and the equipment-check result. Within 24 hours of delivery, record three successes and three changes. Resolve presenter/teleprompter features from the installed renderer or chosen product; no cached app catalog or medical claims are needed.