What Shall We Build Next?

Describe

Describe your task

Refine

Refine the plan

SubAgents

Review all agents

Deploy

Deploy your agent

A) SUBAGENT SUMMARY  
This subagent (DataExtractionSubagent) takes a single product listing URL and extracts four key fields—title, bullets, description, and main image URL—returning them in a JSON-like structure.

B) FINAL TASK OUTPUT  
A JSON structure in the form:  
{  
  "title": "...",  
  "bullets": ["...","..."],  
  "description": "...",  
  "image_url": "..."  
}

C) SUBAGENT INPUT  
{productListingUrl}

E) SUBAGENT TASK SUMMARY  
1) Receive {productListingUrl} as input.  
2) Call skill #226 (Extract Structured Data From 1x URL) with instructions:  
   “Please crawl this URL and extract the product title, bullet points, product description, and main image URL. Return them in JSON format with the fields: 'title', 'bullets', 'description', 'image_url'.”  
3) The skill #226 response will serve as the final subagent output.

F) SILOS  
No silos are needed here because this subagent consists of one primary skill call (skill #226), directly returning the final JSON with the required fields.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent (BrandGuard) takes extracted text and an image URL, checks them for potential trademark or brand references, then returns a concise risk report.

B) FINAL TASK OUTPUT  
[trademark-analysis] → A single text report describing any flagged trademark issues in the text and image.

C) SUBAGENT INPUT  
[extracted-data]  
• Expected structure: { "title": "...", "bullets": ["..."], "description": "...", "image_url": "..." }

D) SUBAGENT OUTPUT  
[trademark-analysis]  
• Text-based: A short summary describing what (if anything) was flagged as trademarked or brand-sensitive.

E) SUBAGENT TASK SUMMARY  
Below is the full sequence of steps (skills) for this subagent. It starts with [extracted-data], identifies possible trademark terms, checks them, analyzes the image, then outputs its concise risk report.

1) (Optional Keyword Extraction)  
   - Skill Used: #223 (Powerful LLM Prompt-to-Text Response)  
   - Input:  
       Prompt containing the text from {title}, {bullets}, and {description}, instructing:  
       “Please read this text and list out any distinctive brand-like terms or phrases that might be potential trademark references or brand names.”  
   - Output: A list of candidate terms or phrases.

2) Text Trademark Check  
   - Skill Used: #224 (Oracle Ask A Question)  
   - Input:  
       For each candidate term from Step 1, ask:  
       “[term] trademark”  
       (Combining multiple terms as needed if the list is lengthy).  
   - Output: A short textual summary of any confirmed or suspicious trademark hits.

3) Image Trademark Check  
   - Skill Used: #176 (Analyze An Image With GPT Vision & Return Text)  
   - Input:  
       The {image_url} from [extracted-data], plus an instruction like:  
       “Check for brand names, logos, or any trademarked elements in this image. Summarize findings.”  
   - Output: Text commentary on visible brand references or suspected trademarked logos.

4) Final Summarization  
   - Skill Used: #223 (Powerful LLM Prompt-to-Text Response)  
   - Input:  
       A prompt that aggregates results from Steps 2 and 3:  
       “Combine these trademark findings for the text and image into a concise risk report.”  
   - Output: [trademark-analysis], describing any flagged terms in text or image.

F) SILOS  
• Silo 1: Text Analysis for Trademarks  
  (Steps 1 & 2)  
• Silo 2: Image Analysis for Trademarks  
  (Step 3)  
• Silo 3: Final Summary  
  (Step 4)

This completes Subagent 2’s flow.

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY:  
This subagent, “TShirtDesignSubagent (DesignGenerator),” generates a brand-new T-shirt design image (PNG/JPG) by combining trademark-safe keywords and style guidelines into a prompt, then using that prompt to create the image.

B) FINAL TASK OUTPUT:  
[new-shirt-design] – A single PNG or JPG image URL representing the newly generated T-shirt design.

C) SUBAGENT INPUT:  
• [extracted-data] – The original listing data (title, bullets, description, image URL).  
• [trademark-analysis] – A concise report detailing any trademarked elements or flagged keywords to avoid.

E) SUBAGENT TASK SUMMARY:  
1. Skill #223 (Powerful LLM Prompt-to-Text Response)  
   – Input:  
     • The extracted text ([extracted-data])  
     • The trademark flags ([trademark-analysis])  
     • Instructions: “Create a T-shirt design prompt that uses only the safe (non-flagged) keywords and style guidelines. Avoid any trademarked references.”  
   – Output: A textual prompt describing the new T-shirt concept.

2. Skill #222 (Make Image (Especially With Text))  
   – Input:  
     • The prompt from Step 1 (the T-shirt concept).  
   – Output: A PNG image URL ([new-shirt-design]) representing the new T-shirt design.

F) SILOS (if any):  
• Silo 1:  
  – Combine trademark-safe keywords and style guidelines using Skill #223 to produce a final T-shirt design prompt.  
• Silo 2:  
  – Generate the new T-shirt design image with Skill #222, based on the prompt from Silo 1.

Final Output: [new-shirt-design] – Ready to be passed along for trademark/compliance re-check and eventual user review.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY:  
This subagent is responsible for taking the extracted listing text (and any trademark flags) and producing a safe, revised title, bullet points, and description that comply with Amazon’s guidelines, then adding a final compliance note.

B) FINAL TASK OUTPUT:  
The final output is a JSON-like object containing:  
{  
  "title": "...",  
  "bullets": ["..."],  
  "description": "...",  
  "compliance_note": "..."  
}

C) SUBAGENT INPUT:  
1) [extracted-data]: The original title, bullets, description, and image URL.  
2) [trademark-analysis]: Includes flagged trademark or brand terms and any specific warnings about trademarked content.

E) SUBAGENT TASK SUMMARY:  
Below is the workflow of action steps (skills) that this subagent will execute to produce the final revised listing text and compliance note:

1) (Preparation / No Skill Needed):  
   - Read in [extracted-data] which contains the original listing text (title, bullets, description).  
   - Read in [trademark-analysis] which contains flagged trademark or brand references.

2) Rewrite Listing Text (Remove Trademarked Terms, Ensure Compliance)  
   - Action: Use #190 - "Write or rewrite text based on instructions."  
   - Input:  
       A prompt specifying that we want a revised title, bullet points, and description that:  
       • Removes or replaces any flagged trademarked terms included in the [trademark-analysis].  
       • Incorporates only safe keywords.  
       • Follows Amazon guidelines for clarity and compliance (e.g., no misleading claims).  
       • Returns the JSON structure containing { "title": "...", "bullets": [...], "description": "..." }.  
   - Output: A newly rewritten listing text in JSON format.

3) Verify Compliance Note  
   - Action: Use #223 - "Powerful LLM Prompt-to-Text Response"  
   - Input:  
       A prompt containing the newly rewritten listing text asking for a short compliance annotation:  
       • “Review the listing text below for compliance with Amazon’s policies. Provide any concerns or confirm that it is compliant. Summarize findings in a single note.”  
   - Output: A short compliance note (text string).

4) Combine into Final Output  
   - The final step merges the rewritten listing text (title, bullets, description) with the compliance note from step 3.  
   - Output format is a JSON-like object:  
       {  
         "title": "...",  
         "bullets": [...],  
         "description": "...",  
         "compliance_note": "..."  
       }

F) SILOS (if any):  
• Silo 1: Rewrite Listing Text (#190)  
• Silo 2: Compliance Check (#223)  
The outputs from Silo 1 feed directly into Silo 2, which then produces the final combined JSON object.