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 visits a single product listing URL (Amazon/Etsy, etc.) and extracts structured data about the product—specifically the product title, bullet points, description, and main image URL.

B) FINAL TASK OUTPUT  
A JSON-like structure, for example:  
{  
  "title": "Sample Title",  
  "bullets": ["Point One", "Point Two"],  
  "description": "Sample long description.",  
  "image_url": "https://domain.com/image.jpg"  
}

C) SUBAGENT INPUT  
{productListingUrl}

E) SUBAGENT TASK SUMMARY  
1) Receive {productListingUrl} as input.  
2) Call Skill #226 - Extract Structured Data From 1x URL  
   • Instruction: “Extract title, bullet points, description, main image URL from the provided page. Output them in JSON-like format.”  
   • Input to Skill #226: {productListingUrl} + the extraction instruction.  
   • Output from Skill #226: A structured text or JSON describing { "title", "bullets", "description", "image_url" }.

3) Return the structured data as the subagent’s final output.

F) SILOS  
This subagent uses only one silo—an extraction silo—performing a single Skill #226 call to collect and format the data.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent (BrandGuard) reviews the extracted product text and image to identify any trademark concerns by checking suspicious text terms via live web search and analyzing the image for brand/logos.

B) FINAL TASK OUTPUT  
[trademark-analysis] – A concise text report summarizing any flagged trademark risks in both the text and the image.

C) SUBAGENT INPUT  
[extracted-data]: A JSON-like structure containing:  
{  
  "title": "...",  
  "bullets": ["..."],  
  "description": "...",  
  "image_url": "..."  
}

E) SUBAGENT TASK SUMMARY  
Below is the detailed step-by-step workflow, in which the numbered actions connect logically from one skill’s output to the next:

1) Extracting Potential Trademark Terms from Text  
   • Use Skill #223 (Powerful LLM Prompt-to-Text Response)  
   INPUT: The combined textual fields from [extracted-data] (title, bullets, description), plus an instruction like:  
   “Analyze this text. Identify individual words or short phrases that could be subject to trademark issues. Return them as a list.”  
   OUTPUT: A list or brief JSON (e.g., ["TermA","TermB","TermC"]). (Call this interim output listOfSuspectTerms)

2) Trademark Check for Each Suspect Term  
   • For each item in listOfSuspectTerms, use Skill #224 (Oracle Ask A Question)  
   INPUT: For example: “[TermA] trademark”  
   OUTPUT: For each query, returns a short result indicating whether the term is likely registered or in use. (Call this brandCheckResults)

3) Image Analysis for Trademarked Content  
   • Use Skill #176 (Analyze An Image With GPT Vision & Return Text)  
   INPUT: The image_url from [extracted-data], with an instruction such as:  
   “Analyze this image. Describe any recognizable brand logos, icons, or text that might be trademarked.”  
   OUTPUT: A textual summary of the image describing any brand references/logos. (Call this imageAnalysis)

4) Summarize Findings into a Concise Report  
   • Use Skill #223 (Powerful LLM Prompt-to-Text Response)  
   INPUT: The brandCheckResults from step 2 plus imageAnalysis from step 3, along with an instruction like:  
   “Combine the following text trademark checks and image analysis into a concise final report (maximum ~200 words) highlighting any trademarks or potential legal concerns.”  
   OUTPUT: A single coherent text response detailing any flagged terms/logos and summarizing that risk.

(This final text is returned as [trademark-analysis].)

F) SILOS  
• Silo 1: Keyword Extraction from the Listing Text (Step 1)  
• Silo 2: Trademark Check via Live Web Queries (Step 2)  
• Silo 3: Image Analysis with GPT Vision (Step 3)  
• Silo 4: Create Final Trademark Report (Step 4)

This completes the trademark analysis subagent flow, yielding a final concise trademark risk report in text form.

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent (DesignGenerator) generates one new trademark-safe T-shirt design image (PNG or JPG) by combining the trademark-safe keywords with style/guidelines, crafting an AI prompt, and then producing the final image.

B) FINAL TASK OUTPUT  
[new-shirt-design]: One newly generated T-shirt design image (PNG/JPG), suitable for use in a print-on-demand listing.

C) SUBAGENT INPUT  
• {extracted-data}: The parsed data from the original listing, e.g. { "title": "...", "bullets": [...], "description": "...", "image_url": "..." }  
• {trademark-analysis}: Short report indicating which keywords or phrases are trademarked and must be avoided, and any image concerns.

D) SUBAGENT TASK SUMMARY  
Step-by-step flow for TSHIRTDESIGNSUBAGENT (DesignGenerator)  
1) (Optional) Summarize or refine the design requirements and safe keywords.  
   - Skill: #223 (Powerful LLM Prompt-to-Text Response)  
   - Input: Subagent receives {extracted-data}, {trademark-analysis}, plus an instruction to produce a short text with only trademark-safe terms and guidelines for the design style.  
   - Output: Cleaned-up text or short bullet list of safe designs/themes.

2) Craft a text prompt specifically for generating a new T-shirt design.  
   - Skill: #223 (Powerful LLM Prompt-to-Text Response)  
   - Input: The refined/safe terms from step 1, plus instructions such as “Draft an AI image-generation prompt that keeps the design distinct from the original and avoids trademarked elements. Emphasize [vintage style / abstract patterns / bright colors / etc.].”  
   - Output: A single text prompt suitable for image-generation (e.g., “Create a vintage-inspired red shirt featuring abstract geometric shapes, no trademark text or logos.”).

3) Generate the new T-shirt design image.  
   - Skill: #222 (Make Image (Especially With Text))  
   - Input: The text prompt from step 2.  
   - Output: A single PNG (or JPG) URL of the newly generated T-shirt design.

4) Return the newly generated T-shirt design.  
   - Final subagent output: [new-shirt-design] (the image URL).

E) SILOS  
Only one main silo is needed here:

• Silo: “Generate T-Shirt Image”  
   1) Summarize safe keywords (#223) → 2) Write final design prompt (#223) → 3) Create T-shirt design (#222) → 4) Output [new-shirt-design].

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY:  
This subagent (ListingCreator) takes in extracted product data plus trademark analysis, then generates brand-new listing text (title, bullet points, description) that avoids trademark issues and complies with Amazon guidelines, returning a final JSON text object plus a short compliance note.

B) FINAL TASK OUTPUT:  
A JSON-like structure containing the revised listing text and a separate compliance note, for example:  
{  
  "title": "...",  
  "bullets": ["..."],  
  "description": "...",  
  "complianceNote": "..."  
}

C) SUBAGENT INPUT:  
1) [extracted-data] → the structured data from Subagent 1 (DataExtractor), which includes:  
   { "title": "...", "bullets": [...], "description": "...", "image_url": "..." }  
2) [trademark-analysis] → the trademark analysis report from Subagent 2 (BrandGuard) indicating flagged terms, safe keywords, and any suspected trademark conflicts.

D) SUBAGENT TASK SUMMARY (Detailed Step-by-Step Flow):

1) INTERPRET TRADEMARK ANALYSIS  
   - Action: Use Skill #223 (Powerful LLM Prompt-to-Text Response).  
   - Input: A prompt instructing the LLM to read [trademark-analysis] and identify which words or phrases are safe to use, which are flagged, and generate a quick list of suggested “safe keywords” for rewriting the listing text.  
   - Output: A text response identifying safe keywords and summarizing trademark concerns.

2) CREATE NEW LISTING TEXT  
   - Action: Use Skill #190 (Write or rewrite text based on instructions).  
   - Input:  
     a) The original listing text (title, bullets, description) from [extracted-data].  
     b) The safe keywords (from Step 1’s text analysis).  
     c) Instructions to remove or replace any flagged terms, ensure clarity, and follow Amazon’s style/quality guidelines.  
   - Output: Draft revised listing text, e.g. { "title": "...", "bullets": [...], "description": "..." } (still not final, pending compliance check).

3) COMPLIANCE CHECK (FINAL CONFIRMATION)  
   - Action: Use Skill #223 (Powerful LLM Prompt-to-Text Response) again.  
   - Input:  
     a) The newly revised listing text from Step 2.  
     b) A request that the LLM confirm any possible trademark or Amazon policy issues, returning a short compliance note.  
   - Output: A text response summarizing whether the new listing is safe and any final recommendations.

4) RETURN FINAL OUTPUT  
   - Combine the revised listing text (title, bullets, description) from Step 2 with the compliance note from Step 3 into one JSON-like structure:  
     {  
       "title": "...",  
       "bullets": ["..."],  
       "description": "...",  
       "complianceNote": "..."  
     }

E) SILOS:  
SILO 1: (Trademark Interpretation)  
• Input: [trademark-analysis]  
• Skill: #223 (Identify safe vs. flagged terms)  
• Output: Safe keywords / flagged terms summary

SILO 2: (Rewrite Listing)  
• Input: [extracted-data] + Safe keywords  
• Skill: #190 (Rewrite text)  
• Output: Draft revised listing text

SILO 3: (Compliance Check)  
• Input: Draft revised listing text  
• Skill: #223 (Compliance confirmation)  
• Output: complianceNote

SILO 4: (Combine & Return)  
• Combine revised listing text + compliance note  
• Final output: [revised-listing-text] JSON-like object with compliance note