popjam-growth-enginelisted

# POPJAM Growth Engine Turn one website URL into tested ad creatives with evidence about *who* converts and *why*. This skill re-implements POPJAM's production pipeline as agent instructions: the same data contracts, the same prompts that carry quality, the same deterministic scoring — but running entirely in this session, persisting to files instead of a database. The loop: **research → audiences → personas → strategy → media → simulate → iterate**. Each phase has a reference file with the distilled production logic; read it when you enter the phase, not before. Phases build on files, so a run can stop anywhere and resume later. ## Campaign state (the contract everything shares) All state lives in one campaign directory — default `growth/<brand-slug>/` under the project root, created at start (ask only if a plausible dir already exists elsewhere): ``` growth/<slug>/ ├── brand.json products.json research.md brief.md # phase 1 ├── audiences/<aud-slug>.json # phase 2 ├── personas/<aud-slug>/<persona-slug>.json # phase 3 ├── strategy/<aud-slug>-angles.json # phase 4 ├── concepts/<concept-id>.json # phase 4 (variants: -v2 + og_id) ├── creatives/<concept-id>/ # phase 5 (real assets only) ├── simulations/<concept-id>/reactions.json, insight.json # phase 6 ├── leaderboard.md # regenerated after each sim round └──