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 (DataExtractionBot) crawls the provided Amazon/Etsy product URL to retrieve title, bullet points, description, and main image URL in a clean, structured format.

B) FINAL TASK OUTPUT  
[extracted-product-details] → A single JSON (or structured text) object containing:  
{  
  "title": "...",  
  "bullets": [ "...", "..." ],  
  "description": "...",  
  "imageURL": "..."  
}

C) SUBAGENT INPUT  
{product-url} → The Amazon/Etsy product page URL to extract data from.

E) SUBAGENT TASK SUMMARY

1) #226 - Extract Structured Data From 1x URL  
   - INPUT:  
     {  
       "url": "{product-url}",  
       "fields": [  
         "title",  
         "bullet points",  
         "description",  
         "main image URL"  
       ]  
     }  
   - OUTPUT:  
     A text-based structured summary of the URL’s page content, containing the requested fields but still in plain text format.

2) #223 - Powerful LLM Prompt-to-Text Response  
   - INPUT:  
     The text output of #226, with instructions to transform the extracted data into JSON with the keys (title, bullets, description, imageURL).  
   - OUTPUT:  
     The final JSON object containing:  
     {  
       "title": "...",  
       "bullets": [ "...", "..." ],  
       "description": "...",  
       "imageURL": "..."  
     }

F) SILOS  
There is only one silo here:  
• Data extraction and conversion to JSON.

The resulting JSON object is then handed off as [extracted-product-details] for use in subsequent subagents.

SubAgent #1 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
“ComplianceCheckerBot examines the text fields from an extracted product listing (title, bullet points, description), identifies any potentially trademarked or policy-violating terms, checks those terms against a trademark database (plus general policy rules), then rewrites or removes them to create a final, safe text output.”

B) FINAL TASK OUTPUT  
[compliant-text-and-keywords]  
This is a single consolidated object or text block that includes:  
• cleaned and safe title/bullets/description, and  
• the final approved keywords (minus trademarked or policy-violating terms).

C) SUBAGENT INPUT  
[extracted-product-details]  
For example,  
{  
  "title": "...",  
  "bullets": ["...", "..."],  
  "description": "...",  
  "imageURL": "..."  
}

D) SUBAGENT TASK SUMMARY  
Below is the detailed flow of actions (skills) that ComplianceCheckerBot will execute. The goal is to produce policy- and trademark-compliant text from the original extracted text.

1) Receive Input ( [extracted-product-details] )  
2) Skill #223 - “Powerful LLM Prompt-to-Text Response”:  
   • Prompt the LLM to scan the provided “title,” “bullets,” “description” for any possibly infringing or policy-violating terms (e.g., brand names, hashtags, references to offensive content).  
   • Output a list/set of “suspicious terms” plus short reasoning for each flagged term.

3) Skill #224 - “Oracle Ask A Question” (Trademark Check):  
   • For each suspicious term found in step 2, query the trademark database or relevant sources using this skill.  
   • Output the confirmation on which terms appear to be trademarked or “high-risk.”

4) Skill #190 - “Write or rewrite text based on instructions”:  
   • Use the original text (“title,” “bullets,” “description”) and the flagged trademark terms from step 3 to rewrite or remove them as needed.  
   • Include any policy compliance changes (e.g., removing or rewriting violent/offensive phrases).  
   • Output the final cleaned text, plus final keywords (if relevant) in one consolidated text block.

5) Subagent returns the result as its “Final Output.”

E) SILOS  
• Silo 1: Identify Suspicious Terms  
  → Input: [extracted-product-details]  
  → Skill #223 to parse text and return suspicious terms.

• Silo 2: Verify Trademark Status  
  → Input: Suspicious terms from Silo 1  
  → Skill #224 to check each term.

• Silo 3: Rewrite Text For Compliance  
  → Input: Original text + results of Silo 2  
  → Skill #190 to remove/replace flagged terms, finalize safe text and keywords.

• Final Output returned: [compliant-text-and-keywords]

SubAgent #2 - Diagram

Expand Diagram

A) SUBAGENT SUMMARY  
This subagent, DesignGeneratorBot, generates a brand-new T-shirt design image in PNG format, ensuring it has a 4500×5400 dimension (300 DPI) and a transparent background based on the cleaned keywords and text coming from the compliance step.

B) FINAL TASK OUTPUT  
[new-tshirt-design-png] → A single PNG file at 300 DPI, sized 4500 × 5400 pixels, with transparent background (ready for Merch by Amazon).

C) SUBAGENT INPUT  
[compliant-text-and-keywords] → Text containing safe words/phrases (already verified as compliant with Amazon’s Merch on Demand policies).

E) SUBAGENT TASK SUMMARY  
Below is the detailed flow (step-by-step actions/skills) for Subagent 3, from receiving [compliant-text-and-keywords] to producing [new-tshirt-design-png]:

1. (Optional) Craft an Image Prompt → Skill #190 (Write or rewrite text based on instructions)  
   • Input: [compliant-text-and-keywords], plus instructions to create a concise image-generation prompt.  
   • Output: [image-generation-prompt-text] (e.g., “Generate a vintage-style geometric red shirt design with no background”).

2. Generate Rough T-Shirt Design Image → Skill #222 (Make Image)  
   • Input: [image-generation-prompt-text] (or [compliant-text-and-keywords] directly if preferred).  
   • Output: [raw-generated-png], a PNG with as little background as possible (the prompt should aim for transparency if feasible).

3. Remove Background (If Not Already Transparent) → External remove-bg API  
   • Input: [raw-generated-png]  
   • Output: [bg-removed-png], a PNG with fully transparent background.

4. Resize Image to 4500 × 5400 → Skill #191 (Resize Image)  
   • Input: [bg-removed-png], with required dimension: 4500 (width) × 5400 (height).  
   • Output: [resized-png]. (Note: This handles pixel dimensions. The next step addresses DPI.)

5. Increase DPI to 300 → External AI image upscaler API  
   • Input: [resized-png]  
   • Output: [final-tshirt-design], a PNG at 300 DPI, 4500 × 5400, transparent background.

6. Final Output  
   • Return [new-tshirt-design-png] = [final-tshirt-design]

F) SILOS  
There are no distinct silos in this subagent’s flow; it proceeds linearly from generating the design to removing background to resizing/upscaling. The final deliverable is a single high-quality PNG.

SubAgent #3 - Diagram

Expand Flow

A) SUBAGENT SUMMARY  
This subagent (ListingTextCreatorBot) takes the cleaned, policy-compliant text/keywords produced by ComplianceCheckerBot and transforms them into an Amazon listing’s final text: title, bullet points, and description.

B) FINAL TASK OUTPUT  
[revised-listing-text] → A single text block containing the final listing details (title, bullet points, description).

C) SUBAGENT INPUT  
[compliant-text-and-keywords] (the consolidated, policy-safe text and set of keywords produced by Subagent 2).

D) SUBAGENT TASK SUMMARY  
• Input: [compliant-text-and-keywords]  
• Action Step:  
   1) #190 - Write or rewrite text based on instructions.  
      - Instruct the LLM to incorporate the safe keywords into a final listing text with the following structure:  
         • Title  
         • Bullet points  
         • Description  
      - Ensure the text meets Amazon’s guidelines (clear, concise, no trademark infringement, no restricted content).  
• Output: [revised-listing-text]

E) SILOS  
No separate silos are needed; there is only one step in which the subagent uses skill #190 to produce the final listing text block.

4 Template & Links

Expand Flow

A) SUBAGENT SUMMARY  
FinalOutputBot is responsible for performing a last-minute check of the newly generated T-shirt design (PNG) and the updated listing text, confirming compliance and trademark safety, and then delivering both as a final package.

B) FINAL TASK OUTPUT  
[final-package] → A bundled delivery containing:  
• The final PNG file (4500 × 5400 px, 300 DPI, transparent background).  
• The final text listing (title, bullet points, description).

C) SUBAGENT INPUT  
[new-tshirt-design-png], [revised-listing-text]

E) SUBAGENT TASK SUMMARY  
1) #224 - Oracle Ask A Question  
   • PURPOSE: Perform a quick final trademark/policy check on the combined new text and design.  
   • INPUT NEEDED: A text prompt to the oracle containing the final listing text and references to the newly generated design, asking “Are there any trademark or content policy issues here?”  
   • OUTPUT: Text indicating the result of the final compliance check.

2) #223 - Powerful LLM Prompt-to-Text Response  
   • PURPOSE: Summarize or collate final references (the PNG file path + the final text-based file) into one cohesive final output. This ensures the user is handed one "final package" containing the link/path to the PNG and the final listing text.  
   • INPUT NEEDED: Prompt that includes the new T-shirt design PNG link and the final text listing, plus any instructions on how to format the final “delivery” to the user.  
   • OUTPUT: [final-package] - A concluding text that details the final PNG URL and the final listing text together in a clear, ready-to-use package.

F) SILOS  
This subagent is brief and does not require separate silos; the two individual skills (#224 → #223) execute in direct sequence to produce [final-package].