start-vibing-stacks 2.29.0 → 2.30.0

package/stacks/_shared/ads-agents/skills/ads-meta/SKILL.md

@@ -0,0 +1,167 @@
+---
+name: ads-meta
+version: 1.6.0
+description: "Meta Ads deep analysis covering Facebook and Instagram advertising. Evaluates 50 checks across Pixel/CAPI health, creative diversity and fatigue, account structure, and audience targeting. Includes Advantage+ assessment. Use when user says Meta Ads, Facebook Ads, Instagram Ads, Advantage+, or Meta campaign."
+user-invokable: false
+---
+
+<!-- Updated: 2026-05-14 | v1.6: Andromeda 2026 playbook (8-12 concepts × 2-3 variations, similarity <40%, 2-3 week fatigue, 20+ tests/month), Advantage+ creative 22% ROAS lift, single-ad-set consolidation -->
+
+# Meta Ads Deep Analysis
+
+## Andromeda AI Engine (2026)
+
+Meta's Andromeda AI engine (global rollout complete Oct 2025) filters ads using
+10,000x more complex retrieval models on the NVIDIA Grace Hopper Superchip. Creative
+diversity is now the #1 performance lever, replacing targeting and bidding as the
+primary optimization surface.
+
+**Similarity Score thresholds (2026 verified):**
+- **>60%** = retrieval suppression (ads collapsed under one Entity ID, compete with each other)
+- **<40%** = recommended target across active assets (treated as distinct candidates)
+- 100 minor variations (color swaps, text tweaks on same template) perform no better than 10 genuinely distinct concepts
+
+**Creative volume reality (2026 data):**
+- Top third of advertisers run ~395 live ads at any time; bottom third runs ~296 (33% gap)
+- Brands testing 20+ new ads/month see 65% higher ROAS than brands testing <10
+- Andromeda-aligned structures deliver 20-35% higher ROAS than legacy structures
+- Advantage+ creative drives 22% ROAS lift over manual creative setups
+- One ad set with 25 diverse creatives outperformed five ad sets with five each by 17% on conversions, 16% on cost
+
+**The 2026 creative system (recommended workflow):**
+1. Build **8-12 conceptually distinct concepts** per campaign (different hooks, formats, emotional triggers, visual treatments)
+2. Produce **2-3 variations per concept** (the variations must differ on visual/audio/structural levels Andromeda actually parses, not just headline swaps)
+3. Refresh on **2-3 week cycle** (fatigue window has shrunk from 6+ weeks pre-Andromeda)
+4. Consolidate into fewer ad sets with more creatives (algorithm performs best when not fragmented)
+5. Mix high-production brand assets with native UGC/founder-led content; neither alone wins
+
+## Process
+
+1. Collect Meta Ads data (Ads Manager export, Events Manager screenshot, EMQ scores)
+2. Read `ads-agents/references/meta-audit.md` for full 50-check audit
+3. Read `ads-agents/references/benchmarks.md` for Meta-specific benchmarks
+4. Read `ads-agents/references/scoring-system.md` for weighted scoring
+5. Evaluate all applicable checks as PASS, WARNING, or FAIL
+6. Calculate Meta Ads Health Score (0-100)
+7. Generate findings report with action plan
+
+## What to Analyze
+
+### Pixel / CAPI Health (30% weight)
+- Meta Pixel installed and firing on all pages
+- Conversions API (CAPI) active (30-40% data loss without it post-iOS 14.5)
+- Event deduplication configured (event_id matching, ≥90% dedup rate)
+- Event Match Quality (EMQ) ≥8.0 for Purchase event
+- All standard events configured (ViewContent, AddToCart, Purchase, Lead)
+- Custom conversions created for non-standard events
+- Aggregated Event Measurement (AEM) configured for iOS
+- Domain verification completed
+- Server-side events include customer_information parameters
+- Pixel fires with correct currency and value parameters
+
+### Creative (30% weight)
+- ≥3 creative formats active (image, video, carousel, collection)
+- ≥5 creatives per ad set baseline; **target 8-12 distinct concepts** per campaign with 2-3 variations each (Andromeda 2026 standard)
+- Creative fatigue detection: CTR drop >10% week-over-week OR >20% over 14 days = FAIL (Andromeda fatigue window: 2-3 weeks, down from 6+)
+- Video creative: 15s max for Stories/Reels, 30s max for Feed
+- UGC/testimonial creative tested AND mixed with polished brand assets (both belong in rotation)
+- Advantage+ Creative enhancements enabled (drives ~22% ROAS lift over manual)
+- Ad copy: headline under 40 chars, primary text under 125 chars
+- Creative refresh cadence: every 2-3 weeks for high-spend (>$5K/month → 5-15 new creatives/week)
+- Creative Similarity Score <40% across active ads (>60% triggers retrieval suppression)
+
+### Account Structure (20% weight)
+- Campaign Budget Optimization (CBO) vs Ad Set Budget (ABO) intentional
+- Campaign consolidation: 1-3 campaigns total recommended
+- Learning phase health: <30% ad sets in "Learning Limited" (FAIL >50%)
+- Budget per ad set: ≥5x target CPA (minimum for learning phase exit)
+- Ad set audience overlap <30% (Audience Overlap tool)
+- Campaign naming conventions consistent and descriptive
+- Advantage+ Sales Campaigns active for e-commerce
+- Simplified campaign structure: 1-3 campaigns total (fewer, larger ad sets preferred)
+
+### Audience & Targeting (20% weight)
+- Prospecting frequency (7-day): <3.0 (WARNING 3-5, FAIL >5)
+- Retargeting frequency (7-day): <8.0 (WARNING 8-12, FAIL >12)
+- Custom Audiences: website visitors, customer lists, engagement
+- Lookalike Audiences: multiple seed sizes tested (1%, 3%, 5%)
+- Advantage+ Audience tested vs manual targeting
+- Interest targeting: broad enough for algorithm optimization
+- Exclusions: purchasers excluded from prospecting, overlap managed
+- Location targeting reviewed for relevance
+
+## Advantage+ Assessment
+
+If Advantage+ features are in use:
+- **Advantage+ Sales Campaigns**: catalog connected, existing customer cap set
+- **Advantage+ Audience**: performance vs manual audience compared
+- **Advantage+ Creative**: enhancements enabled (text, brightness, music)
+- **Advantage+ Placements**: enabled (let Meta optimize placement mix)
+- **Budget allocation**: Advantage+ campaigns getting fair test budget
+
+## Special Ad Categories
+
+If ads are in restricted categories:
+- Special Ad Category declared before campaign creation
+- Targeting restrictions verified (no ZIP, age 18-65+ only, no Lookalike)
+- Creative compliance with category-specific policies
+- Read `ads-agents/references/compliance.md` for full requirements
+
+## EMQ Optimization Guide
+
+| EMQ Score | Status | Action |
+|-----------|--------|--------|
+| 8.0-10.0 | Excellent | Maintain current setup |
+| 6.0-7.9 | Good | Add more customer_information parameters |
+| 4.0-5.9 | Fair | Implement CAPI, improve data quality |
+| <4.0 | Poor | Critical: CAPI + Enhanced Matching required |
+
+Key parameters to maximize EMQ:
+- `em` (email): highest match rate signal
+- `ph` (phone): second highest match signal
+- `fn`, `ln` (first/last name): improves match accuracy
+- `ct`, `st`, `zp` (city, state, zip): geographic matching
+- `external_id`: CRM/user ID for cross-device matching
+
+## Key Thresholds
+
+| Metric | Pass | Warning | Fail |
+|--------|------|---------|------|
+| EMQ (Purchase) | ≥8.0 | 6.0-7.9 | <6.0 |
+| Dedup rate | ≥90% | 70-90% | <70% |
+| CTR | ≥1.0% | 0.5-1.0% | <0.5% |
+| Creative formats | ≥3 | 2 | 1 |
+| Creatives per ad set | ≥5 (target 8-12 distinct concepts × 2-3 variations per campaign) | 3-4 | <3 |
+| Distinct concepts (Andromeda) | ≥10 genuinely distinct | 5-9 | <5 |
+| Creative Similarity Score | <40% | 40-60% | >60% (suppression) |
+| Fatigue: CTR WoW drop | <10% | 10-20% | >20% (kill) |
+| Learning Limited | <30% | 30-50% | >50% |
+| Budget per ad set | ≥5x CPA | 2-5x CPA | <2x CPA |
+
+## Output
+
+### Meta Ads Health Score
+
+```
+Meta Ads Health Score: XX/100 (Grade: X)
+
+Pixel / CAPI Health: XX/100  ████████░░  (30%)
+Creative:            XX/100  ██████████  (30%)
+Account Structure:   XX/100  ███████░░░  (20%)
+Audience:            XX/100  █████░░░░░  (20%)
+```
+
+### Deliverables
+- `META-ADS-REPORT.md`: Full 50-check findings with pass/warning/fail
+- EMQ improvement roadmap
+- Creative fatigue alerts (any creative with CTR declining >20%)
+- Quick Wins sorted by impact
+- Advantage+ adoption recommendations
+
+## Threads Placement
+
+Threads placement GA Jan 2026, 400M+ MAU. Lower CPMs than Feed/Stories.
+Currently ~0.04% of total spend. Emerging channel. Evaluate:
+- Is Threads placement enabled in Advantage+ Placements?
+- Monitor CPM and engagement vs other placements
+- Early-mover advantage for brands with active Threads presence

package/stacks/_shared/ads-agents/skills/ads-microsoft/SKILL.md

@@ -0,0 +1,163 @@
+---
+name: ads-microsoft
+version: 1.6.0
+description: "Microsoft/Bing Ads deep analysis covering search, Performance Max, Audience Network, and Copilot integration. Evaluates 24 checks with focus on Google import validation, unique Microsoft features, and cost advantage assessment. Use when user says Microsoft Ads, Bing Ads, Bing PPC, Copilot ads, or Microsoft campaign."
+user-invokable: false
+---
+
+<!-- Updated: 2026-05-14 | v1.6: Copilot ad lift recalibrated (15-25% verified vs older 73% claim), PMax for Microsoft Advertising 2026 (combines search + Audience Network + Copilot), Copilot 400M+ MAU Q1 2026 -->
+
+
+# Microsoft Ads Deep Analysis
+
+## Process
+
+1. Collect Microsoft Ads data (account export, UET tag status, import results)
+2. Read `ads-agents/references/microsoft-audit.md` for full 24-check audit
+3. Read `ads-agents/references/benchmarks.md` for Microsoft-specific benchmarks
+4. Read `ads-agents/references/scoring-system.md` for weighted scoring
+5. Evaluate all applicable checks as PASS, WARNING, or FAIL
+6. Calculate Microsoft Ads Health Score (0-100)
+7. Generate findings report with action plan
+
+## What to Analyze
+
+### Technical Setup (25% weight)
+- UET tag installed and firing on all pages (MS01)
+- Enhanced conversions enabled (MS02)
+- Google Ads import validated: URLs, extensions, bids, goals (MS03)
+
+### Syndication & Bidding (20% weight)
+- Search partner network reviewed, low-performers excluded (MS04)
+- Audience Network enabled only if testing intentionally (MS05)
+- Bid targets 20-35% lower than Google (CPC advantage) (MS06)
+- PMax: 300 campaigns/account (vs 100 Google), LinkedIn profile data available, no video support
+- Target New Customers enabled for PMax, Beta 2026 (MS07)
+
+### Campaign Structure (20% weight)
+- Campaign structure mirrors Google or follows best practices (MS08)
+- Budget proportional to Bing volume: typically 20-30% of Google (MS09)
+- LinkedIn profile targeting for B2B (unique advantage) (MS10)
+
+### Creative & Extensions (20% weight)
+- RSA: ≥8 headlines, ≥3 descriptions (MS11)
+- Multimedia Ads tested (unique rich format) (MS12)
+- Ad copy optimized for Bing demographics (MS13)
+- Action Extension utilized (unique to Microsoft) (MS19)
+- Filter Link Extension tested (MS20)
+
+### Settings & Performance (15% weight)
+- Copilot chat placement enabled for PMax: 15-25% CTR lift verified 2026 (MS14)
+- Conversion goals configured natively, not relying on imported (MS15)
+- CPC 20-40% lower than Google for same keywords (MS16)
+- CVR comparable to Google, not >50% lower (MS17)
+- Impression share tracked for brand and top terms (MS18)
+
+## Google Import Validation
+
+Most Microsoft Ads accounts start as Google Ads imports. Critical validation:
+
+### What Transfers Correctly
+- Campaign structure and ad groups
+- Keywords and match types
+- RSA headlines and descriptions
+- Basic bid strategies
+
+### What Needs Manual Review
+- **URLs**: verify all landing page URLs are correct post-import
+- **Extensions**: not all Google extensions have Microsoft equivalents
+- **Bid amounts**: should be 20-35% lower (don't import Google bids as-is)
+- **Conversion goals**: re-create natively for better tracking
+- **Audiences**: import may miss segments, verify all are present
+- **Negative keywords**: verify shared negative lists transferred
+
+### Import Schedule
+- Auto-import: useful but review changes monthly
+- **Warning:** Scheduled imports can re-enable paused campaigns. Deactivate scheduled imports after initial setup.
+- Manual import: more control, recommended for large accounts
+- Never import without post-import audit
+
+## Copilot Integration (2026)
+
+Microsoft's AI assistant creates unique ad opportunities. **400M+ MAU on Copilot
+across Microsoft ecosystem as of Q1 2026** (Stackmatix April 2026).
+
+### Copilot Chat Ads
+- Available in Performance Max campaigns
+- **Conversational placements deliver 15-25% higher CTR** vs traditional search ads (verified 2026 partner data); the older "73% lift" figure was early closed-beta and should be treated as anecdotal — use 15-25% as the realistic working range
+- Copilot Checkout launched Jan 2026 (in-chat purchase)
+- Natural language ad delivery (conversational context)
+- Copilot ads show beneath AI responses with Sponsored labels
+- Sponsored responses GA across Copilot since 2025
+
+### How to Evaluate
+- Is Copilot placement enabled? (If not, HIGH priority for PMax)
+- What % of impressions/clicks come from Copilot?
+- CTR/CVR comparison: Copilot vs traditional placements (expect 15-25% lift, not 73%)
+- Ad copy quality: does it read well in conversational context? Leads with specific benefit (not brand name), second-person language, proof points (numbers/outcomes) that give AI context to surface ad appropriately
+
+### PMax for Microsoft Advertising (2026)
+- Combines search + Audience Network + Copilot placements into a single campaign type with automated creative optimization
+- Works best when fed strong first-party conversion data (offline conversions, Enhanced Conversions, CRM imports)
+- LinkedIn profile targeting layered on top remains the unique B2B advantage
+
+## Microsoft-Unique Features
+
+These features are exclusive to Microsoft Ads; evaluate adoption:
+
+| Feature | Description | Priority |
+|---------|-------------|----------|
+| CTV Ads | Netflix, Max, Hulu, Roku, discovery+. 30-second non-skippable. | Medium |
+| Multimedia Ads | Image-rich search ads with visual elements | Medium |
+| Action Extension | CTA button directly in search ad | Medium |
+| Filter Link Extension | Filterable category links in ad | Low |
+| LinkedIn Profile Targeting | Target by company, industry, job function | High (B2B) |
+| Copilot Chat Placement | Ads within Copilot conversations | High |
+| Auto-generated RSA | Enabled by default Jan 2026, 5% CTR increase | Medium |
+| 9:16 Vertical Video | Vertical video ads (Apr 2025), 90-second max duration | Medium |
+
+## Bing Demographic Context
+
+Microsoft Ads reach a distinct audience:
+- Older demographic (35-65+ over-indexed)
+- Higher household income (top 25% income brackets)
+- Desktop-heavy (Windows default browser = Edge = Bing)
+- Enterprise/corporate users (Office 365 integration)
+
+Ad copy optimization for this audience:
+- Professional tone, less casual than Google/Meta
+- Emphasize quality, reliability, premium positioning
+- Desktop-optimized landing pages matter more
+- B2B messaging resonates strongly
+
+## Key Thresholds
+
+| Metric | Pass | Warning | Fail |
+|--------|------|---------|------|
+| CTR (Search) | ≥2.83% | 1.5-2.83% | <1.5% |
+| CPC (Search) | ≤$1.55 | $1.55-2.50 | >$2.50 |
+| CPC vs Google | 20-40% lower | 10-20% lower | Same or higher |
+| CVR vs Google | Within 20% | 20-50% lower | >50% lower |
+| Impression share (brand) | ≥80% | 60-80% | <60% |
+
+## Output
+
+### Microsoft Ads Health Score
+
+```
+Microsoft Ads Health Score: XX/100 (Grade: X)
+
+Technical Setup:   XX/100  ████████░░  (25%)
+Syndication:       XX/100  ██████████  (20%)
+Structure:         XX/100  ███████░░░  (20%)
+Creative:          XX/100  █████░░░░░  (20%)
+Settings:          XX/100  ████████░░  (15%)
+```
+
+### Deliverables
+- `MICROSOFT-ADS-REPORT.md`: Full 24-check findings with pass/warning/fail
+- Google import validation results
+- Copilot integration readiness assessment
+- Cost advantage analysis (CPC savings vs Google)
+- Microsoft-unique feature adoption checklist
+- Quick Wins sorted by impact

package/stacks/_shared/ads-agents/skills/ads-photoshoot/SKILL.md

@@ -0,0 +1,220 @@
+---
+name: ads-photoshoot
+version: 1.5.0
+description: "Product photography enhancement for ad creatives using banana-claude image generation. Takes a product image and generates 5 professional photography styles for ad use: Studio, Floating, Ingredient, In Use, and Lifestyle. Requires banana-claude (v1.4.1+) with nanobanana-mcp. Triggers on: product photo, product photography, photoshoot, enhance product image, product shoot, product photos for ads, generate product photos, studio shot, lifestyle photo."
+user-invokable: false
+---
+
+# Ads Photoshoot: AI Product Photography
+
+Transforms a product image or description into professional ad-ready photography
+in 5 distinct visual styles. Each style generates at two sizes: 1:1 (Meta/LinkedIn)
+and 9:16 (TikTok/Reels/Stories).
+
+## Quick Reference
+
+| Command | What it does |
+|---------|-------------|
+| `/ads photoshoot` | Interactive: ask for product + styles |
+| `/ads photoshoot --styles studio floating` | Generate only selected styles |
+| `/ads photoshoot --product shoe.jpg` | Start with a product image file |
+| `/ads photoshoot --all-platforms` | Generate all 5 sizes per style |
+
+## Environment Setup
+
+Requires banana-claude (v1.4.1+) with nanobanana-mcp configured.
+Run `/banana setup` to configure API key and MCP.
+
+## Process
+
+### Step 1: Collect Product Information
+
+Ask (combine into one message):
+1. **Product image**: Path to product image file (local) OR product URL OR text description
+   > "Provide a product image path (e.g. ./product.jpg), a URL, or describe your product"
+2. **Product description**: What is it? Key features to highlight? (helps prompt quality)
+3. **Styles to generate**: Which of the 5 styles? (default: all 5)
+   - Studio, Floating, Ingredient, In Use, Lifestyle
+4. **Target platforms**: Which platforms will these run on?
+   - Determines output sizes (default: Meta + TikTok → 1:1 + 9:16)
+
+### Step 2: Load Brand Profile (Optional)
+
+Check for `brand-profile.json` in the current directory.
+
+If found, extract for style injection:
+- `colors.primary` → inject into backgrounds and accent elements
+- `aesthetic.mood_keywords` → inject as atmosphere descriptors
+- `target_audience` → use for Lifestyle and In Use context
+- `imagery.forbidden` → exclude from all prompts
+
+If not found, proceed with standard style templates.
+
+### Step 3: Verify banana-claude
+
+Verify banana-claude is installed (run `/banana setup` to check). If not installed,
+display setup instructions and exit.
+
+### Step 4: Construct Prompts per Style
+
+For each selected style, build the prompt using the template + product description + brand DNA.
+
+#### Style 1: Studio
+Clean, e-commerce style product shot.
+
+**Base template:**
+```
+"[product description], professional product photography, clean white seamless
+background, even studio lighting, soft drop shadow, high detail product focus,
+ecommerce style, [brand.colors.primary] subtle accent reflections if applicable,
+top-down or 3/4 angle, no distractions, catalog quality"
+```
+
+**Composition:** Centered, slight 3/4 angle or flat lay.
+**Output sizes:** 1080×1080, 1080×1920
+
+#### Style 2: Floating
+Dramatic levitation effect.
+
+**Base template:**
+```
+"[product description] floating in mid-air, dramatic floating product shot,
+[brand.colors.primary or brand.aesthetic.mood_keywords[0]] gradient background,
+atmospheric shadow below product, levitation effect, product defying gravity,
+clean modern aesthetic, high contrast, striking visual"
+```
+
+**Composition:** Product centered vertically, ample space above and below.
+**Output sizes:** 1080×1080, 1080×1920
+
+#### Style 3: Ingredient
+Flat lay with components.
+
+**Base template:**
+```
+"[product description] centered flat lay, surrounded by its key ingredients
+or materials artfully arranged, top-down overhead view, clean light background,
+natural texture surface, product as hero element, ingredients scattered with
+intentional negative space, editorial food photography style"
+```
+
+**Composition:** Top-down, product in center, ingredients fanning out.
+**Output sizes:** 1080×1080 (optimal for this style). Also generate 9:16 vertical for TikTok/Reels/Stories placements.
+
+#### Style 4: In Use
+Authentic usage context.
+
+**Base template:**
+```
+"person's hands using [product description] in natural context, lifestyle
+photography, focus on product-hand interaction, shallow depth of field,
+warm natural window light, authentic not staged, [brand.target_audience.profession]
+implied context, [brand.aesthetic.mood_keywords] atmosphere"
+```
+
+**Composition:** Hands prominent, product clearly identifiable, background soft-focus.
+**Note:** Hands only; no full face (avoids model release complications).
+**Output sizes:** 1080×1080, 1080×1920
+
+#### Style 5: Lifestyle
+Aspirational full-context shot.
+
+**Base template:**
+```
+"[product description] in aspirational lifestyle scene, [brand.target_audience.age_range]
+demographic implied environment, [brand.target_audience.profession] context,
+[brand.aesthetic.mood_keywords] atmosphere, golden hour or clean natural lighting,
+editorial photography style, [brand.aesthetic.negative_space] composition,
+product clearly visible and prominent"
+```
+
+**Composition:** Environmental context, product as hero element within the scene.
+**Output sizes:** 1080×1080, 1080×1920
+
+### Iterative Refinement
+
+For iterative refinement: if initial generation doesn't match brand expectations, adjust the prompt by specifying: lighting direction, color temperature, background texture, or product angle before regenerating.
+
+### Step 5: Generate Images
+
+**Domain mode selection per style:**
+- Use banana **Product** mode for Studio, Floating, and Ingredient styles
+- Use banana **Editorial** mode for In Use and Lifestyle styles
+- Set resolution to 2K (default) for all generations
+
+**Aspect ratio setup:** Use banana MCP `set_aspect_ratio` before each generation:
+- For 1080x1080: set ratio to 1:1
+- For 1080x1920: set ratio to 9:16
+
+For each style x size combination, use `/banana generate` with the constructed
+prompt, selected domain mode, and correct aspect ratio. Save output to
+`./product-photos/[style]/[product-slug]-[style]-[WxH].png`.
+
+Track results. If a generation fails, retry once with a simplified prompt.
+
+### Step 6: Organize and Report
+
+**Output directory structure:**
+```
+./product-photos/
+  studio/
+    product-studio-1080x1080.png
+    product-studio-1080x1920.png
+  floating/
+    product-floating-1080x1080.png
+    product-floating-1080x1920.png
+  ingredient/
+    product-ingredient-1080x1080.png
+    product-ingredient-1080x1920.png
+  in-use/
+    product-in-use-1080x1080.png
+    product-in-use-1080x1920.png
+  lifestyle/
+    product-lifestyle-1080x1080.png
+    product-lifestyle-1080x1920.png
+```
+
+**Summary:**
+```
+✓ Product photos generated: [N] images
+
+  Studio:     ./product-photos/studio/ (2 sizes)
+  Floating:   ./product-photos/floating/ (2 sizes)
+  Ingredient: ./product-photos/ingredient/ (2 sizes)
+  In Use:     ./product-photos/in-use/ (2 sizes)
+  Lifestyle:  ./product-photos/lifestyle/ (2 sizes)
+
+  Cost: see .banana/costs.json for total spend
+
+  Best for:
+  • Meta Feed → Studio (4:5) or Lifestyle (4:5)
+  • TikTok/Reels → Floating (9:16) or In Use (9:16)
+  • LinkedIn → Studio (1:1) or Lifestyle (1:1)
+  • Google PMax → Studio (1:1); crop to 1.91:1 after
+
+  Run `/ads generate` to use these in a full campaign.
+```
+
+## Cost Estimate
+
+Before generating, show:
+- Number of styles selected x 2 sizes = total images
+- Estimated cost based on banana pricing tiers
+- If >$0.50, ask for confirmation
+
+## Platform Recommendations
+
+| Style | Best Platforms | Rationale |
+|-------|---------------|-----------|
+| Studio | Meta Feed, LinkedIn, Google PMax | Universal, clean, platform-safe |
+| Floating | Meta Reels, TikTok, Stories | High visual impact on vertical placements |
+| Ingredient | Meta Feed, Pinterest | Works best as square; tells product story |
+| In Use | TikTok, Meta Reels, Stories | Authentic, native-feeling content |
+| Lifestyle | All platforms | Aspirational, broad audience appeal |
+
+## Reference Files
+
+- `ads-agents/references/image-providers.md`: API setup and pricing
+- `ads-agents/references/brand-dna-template.md`: Brand injection schema
+- `ads-agents/references/meta-creative-specs.md`: Safe zone for 9:16
+- `ads-agents/references/tiktok-creative-specs.md`: Safe zone constraints