meta_ads_extractorlisted

# Meta Ads Library Asset Extraction Extract actual ad creative images and videos from Meta's Ad Library via browser automation. > **Portability note:** delivery via the `message` tool (`action=send`) and the Telegram size > limits below assume the OmegaOS VPS bridge. Off-VPS, drop assets into the run's output > directory instead and skip the `message` send step — extraction logic is unchanged. ## Dynamic Workflow orchestration Extraction fans out cleanly over independent **units** = each advertiser (and, within one advertiser, each lazy-loaded batch / each asset). Use this when extracting from multiple advertisers or a large swipe file; for a single small page, run linearly. 1. **Plan** — resolve the unit list: list advertisers/pages requested + the per-advertiser target (top-N, media type, country). One unit = one `view_all_page_id` extraction. 2. **Parallel fan-out** — per advertiser: resolve delegate_page ID -> open filtered Ad Library -> extract all media + CTA URLs from DOM (steps 1-4) -> download only the requested assets (step 5). File-disjoint by advertiser slug, so units never collide (R-SCOPE: one writer per output file). 3. **Adversarial verify (2-of-3)** — accept each unit only when at least 2 of 3 independent checks pass: (a) every requested file exists and is non-zero on disk; (b) dimensions/aspect ratio captured via `sips`/`identify`/`ffprobe` and written to metadata JSON; (c) DOM order matches the requested rank (e.g. `images[0]` = to