trend-ad-composer

# Trend Ad Composer Generate ad concept batches from real signals and submit them to the ingest API. ## Inputs to gather before generation - Offer scope: - Generic business ad concepts, or - Specific campaign tied to upcoming/past training session - Target audience slice: - Marketers - Entrepreneurs/founders - AI consultants/agencies - Batch details: - `client_id` - `batch_id` - Platform mix: - facebook / instagram / linkedin (at least 3 variations per batch) ## Research workflow 1) Check current opportunity signals - Use web search for trend demand and pain-point language - Pull at least 5-8 sources total across: - search trends/news - social/community pain points - competitor/ad messaging signals 2) Review Reinventing.ai session context - Review upcoming webinar and selected past sessions at: - https://www.reinventing.ai - Extract positioning angles: - outcome promise - urgency/time sensitivity - who benefits most 3) Convert signals into ad angles - Prioritize angles with clear economic outcomes: - hours saved - revenue potential - faster launch cadence - reduced execution overhead ## Output requirements Always produce JSON matching the schema expected by the app endpoint. Required top-level keys: - `client_id` - `batch_id` - `generated_at` - `research_context` - `ad_variations` - `generation_metadata` Variation quality rules: - 3-6 variations minimum - Mix broad + niche angles - Include `reasoning` tied to research signals - Include strong image prompts suitable for generation - Include targeting notes and `estimated_hook_strength` - Use platform-appropriate aspect ratios (avoid defaulting to wide) Aspect ratio defaults by platform (preferred): - instagram_feed: `4:5` (1080x1350 style) - instagram_square: `1:1` - instagram_story/reels: `9:16` - facebook_feed: `4:5` - facebook_landscape: `16:9` only when explicitly requested - linkedin_feed_image: `4:5` (or `1:1` secondary) - linkedin_landscape_ad: `16:9` only for specific ad placement Visual creative guidance (mandatory in every `image_prompt`): - Specify composition clearly: - subject placement (left/right/center) - foreground + background elements - depth of field / layering - Include ad-friendly layout intent: - “leave clean negative space for optional headline overlay” - Define emotional tone: - confidence, relief, momentum, authority, etc. - Define style and realism level: - corporate photography, cinematic lifestyle, flat illustration, UI mockup, etc. - Define color direction: - contrast, brand-safe palette, high legibility - Explicitly request: no embedded text, logos, or watermarks ## Copy style constraints - Value-first hooks - Short, specific claims - Avoid generic hype and inflated promises - Prefer concrete outcomes and implementation language - For LinkedIn concepts, no outbound links in body copy ## Endpoint submission Use the helper script: ```bash set -a; source /root/.openclaw/credentials/ads-ingest.env; set +a python3 skills/trend-ad-composer/scripts/submit_ads.py --json-file /path/to/batch.json ``` Environment variables required: - `ADS_INGEST_ENDPOINT` - `ADS_INGEST_API_KEY` ## Optional references - Schema reference: `references/ad-schema.example.json`