build-audiencelisted

# Build Audience ## Purpose Populate Compartment 2 (Audience Intelligence System — the heaviest compartment at 18% weight) to ≥70% completeness. Produces the **Audience Intelligence Document** — a full ICP with demographics, psychographics, voice-of-viewer (20+ verbatim quotes), awareness-level distribution, limiting belief diagnosis, objection library, and behavioral patterns. This is the single most load-bearing artifact in the OS — every downstream skill (script writing, hook engineering, title strategy, thumbnail design, VSL, lead-magnet, sales script) reads from this compartment. If this is wrong or generic, every downstream output is corrupted. ## Decision Logic Audience research on YouTube is structurally different from generic ICP work because the channel produces 50–500 videos per audience archetype, and the audience self-selects through retention curves. The audience is partially **revealed by the algorithm** rather than fully designed by the creator. So the audience profile is written iteratively: the creator's hypothesis → first cohort of videos → retention + comment data → refined audience profile. This skill produces the **first-draft hypothesis** rich enough to enable production, with explicit gaps marked for cycle-2 refinement after early publish data arrives. The VOC (voice-of-viewer) extraction is non-negotiable. 20+ verbatim quotes minimum, sourced from real audience surface (comments, DMs, sales calls, surveys, Reddit threads, adjacent-channel comments)