What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
“Insight & Copy Builder” researches the client’s competitive ad landscape and then writes a structured pack containing (1) a markdown competitor-insight report and (2) a complete set of ready-to-use ad-copy variations, wrapped up with a small manifest file.

B) FINAL TASK OUTPUT  
[insight-copy-bundle] → a folder that contains  
• research/competitor_summary.md (≈1 000–2 500 words, markdown)  
• copy/ad_copy.json (valid JSON with: 5 headlines, 5 sub-headlines, 3 short-form bodies, 3 long-form bodies)  
• insight_copy_manifest.txt (plain-text manifest listing file names, sizes, and a one-line description for each)

C) SUBAGENT INPUT  
{business-name} (string)  
{industry} (string)  
{target-audience} (string)  
{product-description} (string)

E) SUBAGENT TASK SUMMARY

{business-name, industry, target-audience, product-description}  
  → #218 Brainstorm Related Keywords (creates keywords_list.txt – 8-12 highly-relevant search terms)  
  → #224 Oracle Ask A Question (uses the keywords + inputs to run 5 live searches such as  
   “Show top Facebook/Instagram ads for {industry} targeting {target-audience}”,  
   “Best-performing Google text ads for {product-description}”, etc.) – returns raw_ads_research.txt  
  → #223 Powerful LLM (ingests raw_ads_research.txt and extracts: primary angles, hooks, offer types, CTA styles, common creative motifs; outputs competitor_summary.md placed in /research)  
  → #185 Write Text (uses competitor_summary.md + full user inputs to generate ad_copy.json with exact schema:  
   { “headlines”:[], “sub_headlines”:[], “short_form_copy”:[], “long_form_copy”:[] }) – saved in /copy  
  → #223 Powerful LLM (compiles insight_copy_manifest.txt that lists each produced file, its purpose, and byte size)  
  ⇒ Return bundle as [insight-copy-bundle]

F) SILOS

SILO 1 – Competitive Research  
Inputs → #218 → #224 → #223 → competitor_summary.md

SILO 2 – Copy Production & Packaging  
competitor_summary.md + original inputs → #185 → ad_copy.json  
ad_copy.json + competitor_summary.md → #223 → insight_copy_manifest.txt  
→ aggregate three files as [insight-copy-bundle]

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
“Visual Creative Designer” turns the strategic copy/insights from Sub-Agent 1 into three-to-five scroll-stopping static ad images that already contain the chosen headline text and follow the visual tropes of top-performing ads in the niche.

B) FINAL TASK OUTPUT  
[image-creatives] → a folder named images/ that contains  
• creative_01.png … creative_0X.png (3-5 PNGs, default 1080 × 1350, ≤1 MB each)  
• images_manifest.txt (UTF-8 text file ≈ 1-2 KB that documents prompt, headline, angle, alt-text & recommended placement for every image)

C) SUBAGENT INPUT  
{business-name} – string  
[insight-copy-bundle] – folder containing  
 • research/competitor_summary.md  
 • copy/ad_copy.json  
 • insight_copy_manifest.txt

E) SUBAGENT TASK SUMMARY

INPUTS → SKILL CHAIN → OUTPUTS

1. Generate Creative Brief  
   [insight-copy-bundle] → #223 Powerful LLM  
   Prompt:  
   “Analyse ad_copy.json (headlines, sub-headlines, angles) + competitor_summary.md.  
   • Select the 5 strongest headline/angle combos for static ads aimed at the target audience.  
   • For each combo output JSON with keys: filename (‘creative_0X.png’), headline_text, optional_sub_text, call_to_action, visual_style (colours, mood, image type), recommended_aspect_ratio (‘1080×1350’ default), and short_rationale (≤20 words).  
   Return array length 3-5.”  
   Output: creative_brief.json (saved to temp memory)

2. Create Images (loop once per entry in creative_brief.json)  
   For each item (up to 5):  
   a. #222 Make Image (Especially With Text)  
      Input:  
      “Generate a [recommended_aspect_ratio] PNG ad for ‘{business-name}’.  
      Headline: “[headline_text]” (large, bold).  
      Sub-text/CTA: “[optional_sub_text or call_to_action]”.  
      Style: [visual_style] (reflect trending ads in the niche).  
      Follow brand-safe colours, high contrast, no watermarks.”  
      Output: PNG URL → saved as images/creative_0X.png

3. Build Manifest  
   (image URLs + creative_brief.json) → #223 Powerful LLM  
   Prompt:  
   “Combine the list of generated image URLs with creative_brief.json.  
   Produce images_manifest.txt containing, for each creative:  
   • filename  
   • headline_text  
   • short_rationale (the ‘angle’)  
   • alt_text (≤15 words, ADA-compliant)  
   • recommended_platform (‘FB Feed’, ‘IG Story’, etc.)  
   • source_prompt_snippet (first 120 chars of the #222 prompt).  
   Separate entries with ‘---’.”  
   Output: images_manifest.txt

F) SILOS

SILO 1 – BRIEF CREATION  
 Input: [insight-copy-bundle] → #223 → creative_brief.json

SILO 2 – IMAGE GENERATION  
 Loop 3-5× → #222 Make Image → images/creative_0X.png

SILO 3 – MANIFEST  
 Inputs: creative_brief.json + image URLs → #223 → images_manifest.txt

Final deliverable from this sub-agent: [image-creatives] folder containing all PNGs plus images_manifest.txt, ready for Sub-Agent 3 or 4.

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
“Video Ad Producer” turns the insight/copy bundle into a fully-finished 15-30 sec, 9:16 vertical video ad (script ➜ voice-over ➜ text-slide base video ➜ light B-roll overlay ➜ manifest), but only runs when {include-video}=true.

B) FINAL TASK OUTPUT  
[video-ad-assets] folder containing:  
• video_script.txt   (UTF-8 text, ≤200 words)  
• voiceover.mp3      (44.1 kHz, mono/stereo, ≤30 sec)  
• storyboard.json    (JSON array of scenes with on-screen text & visual notes)  
• video_ad.mp4       (MP4, H.264, 1080 × 1920, 15-30 sec, ≤20 MB)  
• video_manifest.txt (plain text summary of all above files)

C) SUBAGENT INPUT  
Required  
{business-name}                – e.g. “SunnySide Café”  
{include-video}                – Boolean (“true” → run, “false” → skip)  
[insight-copy-bundle]          – contains competitor_summary.md & ad_copy.json  
(If {include-video}=false the sub-agent exits immediately and returns nothing.)

E) SUBAGENT TASK SUMMARY  
(Executed only when {include-video}=true)

1. Build storyboard from the winning angles  
   [insight-copy-bundle] ➜ #223 Powerful LLM  
   – Prompt: “Using ad_copy.json headlines & angles, craft a 3-5-scene storyboard (JSON) for a 15-30 sec vertical ad for {business-name}. Each scene needs: ‘time_range’, ‘voiceover’, ‘text_on_screen’, ‘visual_suggestion’.”  
   ➜ OUTPUT: storyboard.json

2. Write concise voice-over script  
   storyboard.json ➜ #171 Write Voice-Over Script  
   – Instructions: “Convert storyboard into 1 continuous script (≤200 words, energetic, conversational).”  
   ➜ OUTPUT: video_script.txt

3. Generate voice-over audio  
   video_script.txt ➜ #170 Turn Script Into Voice-Over MP3  
   ➜ OUTPUT: voiceover.mp3

4. Auto-create time-stamped transcription  
   voiceover.mp3 ➜ #198 Get Transcription Of MP3  
   ➜ OUTPUT: transcription_with_timings.txt  (internal helper file)

5. Produce vertical text-slide base video  
   voiceover.mp3 + transcription_with_timings.txt ➜ #201 Generate Vertical Text Slide Video From MP3 & Transcription  
   ➜ OUTPUT: base_text_video.mp4

6. Fetch 3–5 relevant royalty-free vertical B-roll clips  
   key phrase = first keyword in ad_copy.json headline list ➜ #205 Get Royalty-Free VERTICAL Videos  
   ➜ OUTPUT: b_roll_list.txt  (internal helper list of 5 MP4 URLs with descriptions)

7. Overlay B-roll onto base video (light visuals)  
   base_text_video.mp4 + b_roll_list.txt + transcription_with_timings.txt ➜ #199 Add Images & Videos On Top Of Existing MP4  
   – Params: “Insert 2-3 of the most relevant clips at logical breaks; keep captions visible.”  
   ➜ OUTPUT: video_ad.mp4  (final video)

8. Create manifest of delivered assets  
   List: video_script.txt, voiceover.mp3, storyboard.json, video_ad.mp4 ➜ #223 Powerful LLM  
   – Prompt: “Build video_manifest.txt: file names, durations, sizes, purpose, suggested usage.”  
   ➜ OUTPUT: video_manifest.txt

Final subagent output: [video-ad-assets] = {video_script.txt, voiceover.mp3, storyboard.json, video_ad.mp4, video_manifest.txt}

F) SILOS

SILO 1 – PRE-PRODUCTION  
#223 (storyboard.json) → #171 (video_script.txt)

SILO 2 – AUDIO  
#170 (voiceover.mp3) → #198 (transcription)

SILO 3 – VIDEO  
#201 (base_text_video.mp4) → #205 (B-roll list) → #199 (video_ad.mp4)

SILO 4 – PACKAGING  
#223 (video_manifest.txt) ➜ collate all deliverables into [video-ad-assets]

The sequence ensures every skill’s output matches the next skill’s required input and ends with a ready-to-publish vertical video ad plus all supporting files.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
“Deployment & Packaging Assistant” converts all upstream assets (copy, images, optional video) into ready-to-publish social posts, a concise launch guide, and finally bundles the whole project into a single, well-structured .zip file plus a delivery manifest.

B) FINAL TASK OUTPUT  
[packaged-mini-ad-campaign] →  
• mini_ad_campaign_{business-name}.zip (root contains ⤵)  
   └─ /copy/competitor_summary.md  
   └─ /copy/ad_copy.json  
   └─ /creatives/images/creative_01.png …  
   └─ /creatives/video/video_ad.mp4  (if supplied)  
   └─ /deployment/captions_and_hashtags.json  
   └─ /deployment/quick_start_guide.md  
• delivery_manifest.txt (plain-text inventory & checksum list)

C) SUBAGENT INPUT  
{business-name} (string)  
[insight-copy-bundle]  → competitor_summary.md, ad_copy.json  
[image-creatives]      → 3-5 PNGs + images_manifest.txt  
[video-ad-assets]      → (optional) video_ad.mp4, video_script.txt, voiceover.mp3, video_manifest.txt

E) SUBAGENT TASK SUMMARY  
1. Caption / Hashtag generation  
   [insight-copy-bundle] + {business-name}  
      → #186 Create 2-10 Short MicroCopy   
         • INPUT: short brief compiled from ad_copy.json (angles, USP)  
         • OUTPUT: raw_captions.txt  (≈6 punchy captions)  
      → #223 Powerful LLM Prompt-to-Text Response  
         • INPUT: raw_captions.txt + instruction “add 5-8 relevant hashtags for each caption & output valid JSON array with {caption,hashtags} objects.”  
         • OUTPUT: captions_and_hashtags.json

2. Quick-start deployment guide  
   → #185 Write Text  
     • INPUT: structured prompt containing:  
         – overview of assets (file names from manifests)  
         – recommended ad placements (FB, IG, TikTok, etc.)  
         – how to A/B test headlines & creatives  
         – upload specs + spend suggestions  
     • OUTPUT: quick_start_guide.md

3. Package all assets  
   a. Internal “Assemble Folder Tree” (implicit)  
      • copies upstream assets into /copy, /creatives/images, /creatives/video, /deployment  
   b. Internal Zip & Save  
      • INPUT: root folder path “mini_ad_campaign_{business-name}/”  
      • OUTPUT: mini_ad_campaign_{business-name}.zip

4. Delivery manifest  
   → #223 Powerful LLM Prompt-to-Text Response  
     • INPUT: directory listing of .zip contents + checksum values  
     • OUTPUT: delivery_manifest.txt (human-readable inventory, sizes, optional video flag)

5. Return both mini_ad_campaign_{business-name}.zip and delivery_manifest.txt as [packaged-mini-ad-campaign].

F) SILOS

SILO 1 – Captions & Hashtags  
   #186 → raw_captions.txt  
   #223 → captions_and_hashtags.json

SILO 2 – Quick-Start Guide  
   #185 → quick_start_guide.md

SILO 3 – Packaging & Manifest  
   Internal assemble → Zip & Save → #223 → final outputs