What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
"Listing Data Extractor" takes a single URL (Amazon or Etsy) and returns the product’s original title, bullet points, description, and main image URL as one structured output.

B) FINAL TASK OUTPUT  
[extracted-listing-data] → A structured text (JSON-like) containing:  
• title (string)  
• bullet_points (array of strings)  
• description (string)  
• main_image_url (string)

C) SUBAGENT INPUT  
{listing-url}

E) SUBAGENT TASK SUMMARY  
1) Determine if {listing-url} points to Amazon or another site.  
   - Use Skill #223 (Powerful LLM Prompt-to-Text Response) with a small prompt such as:  
     “Check this URL: {listing-url}. Is it an Amazon domain (yes/no)?”  
2) If the domain is Amazon → Call Skill #225 (Crawl 1x Amazon URL) to extract structured data from the product page.  
   Else → Call Skill #226 (Extract Structured Data from 1x URL) for Etsy or other domain pages.  
3) Take the textual summary output from Step 2 and transform it into a concise JSON-like structure (title, bullet_points, description, main_image_url).  
   - Use Skill #223 again with a prompt:  
     “Convert the following text about the product into JSON keys [title, bullet_points, description, main_image_url]. Return them as valid JSON.”

The final text returned from Step 3 is the subagent’s single structured output.

F) SILOS  
No complex silos are needed; “Listing Data Extractor” simply performs the steps above in sequence to produce [extracted-listing-data].

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent analyzes the original shirt’s main image (colors, style, pattern) and the listing text (title, bullet points, description) to produce:  
• A descriptive summary of the shirt design.  
• A “safe” keyword list, stripped of any potentially trademarked/infringing terms.

B) FINAL TASK OUTPUT  
[safe-keywords-and-image-description] → A single structured data object containing:  
1) "ImageDescription": text summarizing the shirt’s design attributes.  
2) "SafeKeywords": an approved list of keywords free from trademarked/infringing terms.

C) SUBAGENT INPUT  
[extracted-listing-data] → This is the structured output from Subagent 1, which includes:  
• Original title, bullet points, and description.  
• Main image URL (the first image from the listing).

E) SUBAGENT TASK SUMMARY  
Below is the step-by-step sequence of skills used to produce the final output [safe-keywords-and-image-description]:

1) Take the main image URL from [extracted-listing-data]  
   └ Use Skill #176 (Analyze An Image With GPT Vision & Return Text)  
      – INPUT: the main image URL plus an instruction like “Describe the shirt’s style, colors, and any patterns.”  
      – OUTPUT: "ImageDescription" (text summarizing visible design elements).

2) Parse the listing text (title, bullet points, description) from [extracted-listing-data], identify possible trademarked/infringing terms, and remove them.  
   └ Use Skill #223 (Powerful LLM Prompt-to-Text Response)  
      – INPUT: a prompt instructing the LLM to extract keywords from the listing text, check for potential trademark issues, and produce a final curated list of safe keywords.  
      – OUTPUT: "SafeKeywords" (a text-based list/string of words that are free from trademarked or infringing terms).

3) Combine “ImageDescription” and “SafeKeywords” into one unified structured output.  
   └ Use Skill #223 (Powerful LLM Prompt-to-Text Response) again  
      – INPUT: a prompt to merge the two elements (“ImageDescription” and “SafeKeywords”) into a single JSON-like object.  
      – OUTPUT: [safe-keywords-and-image-description], containing both fields.

F) SILOS  
No internal silos or additional subdivisions are required here. The three sequential steps shown above comprehensively fulfill the subagent’s role.

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
"Design Generator & Processor" creates two brand-new T-shirt designs (with transparent backgrounds) at 300 DPI and 4500×5400px using the safe keywords and style descriptors from Subagent 2, ensuring final PNG outputs.

B) FINAL TASK OUTPUT  
Two high-resolution PNG images, each with:  
• Transparent background  
• 300 DPI  
• 4500×5400 pixels

C) SUBAGENT INPUT  
[safe-keywords-and-image-description]  
This is a structured object containing:  
1) A safe list of non-trademarked keywords.  
2) A descriptive summary of the original shirt design (colors, style, and patterns).

E) SUBAGENT TASK SUMMARY  
Below is the step-by-step workflow (one silo) that will generate two distinct design variations.

1) (Optional) Generate Distinct Prompts for Each Variation  
   → Skill Used: #223 (Powerful LLM Prompt-to-Text Response)  
   INPUT:  
       [safe-keywords-and-image-description] with an instruction (e.g., “Create two distinct creative prompts for a new T-shirt design based on these safe keywords and style description. Each prompt must request a transparent background.”)  
   OUTPUT:  
       Two short text prompts (Prompt A, Prompt B) suitable for image generation that incorporate the safe keywords and style details.

2) Generate T-Shirt Design Variation #1  
   → Skill Used: #222 (Make Image (Especially With Text))  
   INPUT:  
       Prompt A (e.g., “Design a T-shirt graphic with [keywords], style [image-description], ensure transparent background...”)  
   OUTPUT:  
       PNG image URL (variation1_raw.png).

3) Resize Variation #1 to Required Specs  
   → Skill Used: #191 (Resize Image)  
   INPUT:  
       variation1_raw.png + instructions to resize to 4500 (width) × 5400 (height), 300 DPI  
   OUTPUT:  
       variation1_final.png (transparent, 4500×5400, 300 DPI).

4) Generate T-Shirt Design Variation #2  
   → Skill Used: #222 (Make Image (Especially With Text))  
   INPUT:  
       Prompt B (e.g., “Design a different T-shirt graphic with [keywords], style [image-description], ensure transparent background...”)  
   OUTPUT:  
       PNG image URL (variation2_raw.png).

5) Resize Variation #2 to Required Specs  
   → Skill Used: #191 (Resize Image)  
   INPUT:  
       variation2_raw.png + instructions to resize to 4500 (width) × 5400 (height), 300 DPI  
   OUTPUT:  
       variation2_final.png (transparent, 4500×5400, 300 DPI).

6) Final Subagent Output  
   [generated-tshirt-image], comprising:  
   • variation1_final.png  
   • variation2_final.png

F) SILOS  
Because this subagent’s objective is straightforward (two image variations in the specified format), the entire workflow is in a single silo called “Design Generation & Processing.” The above steps produce the final pair of PNG images ready for downstream subagents or final delivery.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
"Listing Text Creator" takes the safe keywords and design description from Subagent 2, then drafts and finalizes new listing text (title, bullet points, description) that complies with Amazon guidelines (length limits, no trademarked terms).

B) FINAL TASK OUTPUT  
[final-listing-text]: A single text file (in .txt or CSV format) containing:  
• Title (≤60 characters)  
• 2 bullet points (each ≤200 characters)  
• Description (≤1000 characters),  
All free of infringing or trademarked terms.

C) SUBAGENT INPUT  
[safe-keywords-and-image-description]: Structured object from Subagent 2, containing non-trademarked keywords and a summary of the original shirt design’s style/colors.

E) SUBAGENT TASK SUMMARY

1) Input: [safe-keywords-and-image-description]  
2) SKILL #190 - "Write or rewrite text based on instructions":  
   • Draft the Amazon-compliant listing text (title, bullet points, description).  
   • Incorporate:  
     – Non-trademarked keywords from [safe-keywords-and-image-description]  
     – Style references (if needed)  
     – Length requirements:  
       → Title ≤ 60 characters  
       → 2 bullet points, each ≤ 200 characters  
       → Description ≤ 1000 characters  
     – Exclude any trademarked terms (avoid words flagged in prior steps, e.g. “T-shirt,” “holiday” references, etc.)  
   • Output a single cohesive draft listing text.

3) SKILL #223 - "Powerful LLM Prompt-to-Text Response":  
   • Input: Draft listing text from step (2).  
   • Check for any remaining trademarked terms or policy violations.  
   • If issues remain, remove or replace them.  
   • Finalize the listing text.  
   • Produce [final-listing-text] in a single text output (to be saved as .txt or .csv).

F) SILOS (none needed beyond these two steps)

The subagent completes by outputting [final-listing-text], containing the final title, bullet points, and description, properly formatted and free of any trademark or policy violations.