> ## Knowledge Base Index > Fetch the complete knowledge base index at: https://help.ubuyfirst.com/sitemap.xml > Use this file to discover available pages before exploring further. > Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash). # Building Your AI Analysis ## Create and refine prompts, display templates, and custom columns Table Of Contents * [Create and refine prompts, display templates, and custom columns](#0-toc-title) * [What You’re Creating](#1-toc-title) * [Before You Start](#2-toc-title) * [Step 1: Set Up Your AI Project](#3-toc-title) * [Option A: ChatGPT (Easiest)](#4-toc-title) * [Option B: Claude](#5-toc-title) * [Option C: Gemini](#6-toc-title) * [What’s in the Knowledge Files?](#7-toc-title) * [Step 2: Generate Your First Prompt & Template](#8-toc-title) * [Start the Conversation](#9-toc-title) * [Describe Your Mental Process](#10-toc-title) * [Example Conversation](#11-toc-title) * [What Your AI Assistant Will Do](#12-toc-title) * [Step 3: Load Into SKU Manager](#13-toc-title) * [Save the Prompt](#14-toc-title) * [Save the Display Template](#15-toc-title) * [Step 4: Create Your Filter (in uBuyFirst)](#16-toc-title) * [Step 5: Test It](#17-toc-title) * [Improving Your Prompt Over Time](#18-toc-title) * [When AI Gets Something Wrong](#19-toc-title) * [Adding Features Later](#20-toc-title) * [Quick Reference](#21-toc-title) * [Adding Custom Columns (Optional)](#22-toc-title) * [Why Use Both?](#23-toc-title) * [How to Add Columns](#24-toc-title) * [Columns Enable Additional Filters](#25-toc-title) * [Include Reasoning in Your Prompts](#26-toc-title) * [What Good Output Looks Like](#27-toc-title) * [Tips for Success](#28-toc-title) --- ## What You’re Creating You’ll be building two components **outside of uBuyFirst** using an AI assistant, then bringing them back into the application: | Component | What It Does | Where It Lives in SKU Manager | | --- | --- | --- | | Prompt | Instructions telling the AI how to analyze listings | AI Prompt tab | | Display Template | HTML that formats results in External Panel | AI Display Template tab | The prompt outputs JSON data → The template displays it visually. Once set up, your AI project becomes your ongoing workspace — you’ll come back here to make enhancements, add features, and fix issues using feedback traces. --- ## Before You Start ☑️ **SKU Manager configured** — Python installed, API key set up, Fields to Send profile created (see Quick Start Guide) ☑️ **AI assistant account** — Pick one: * **ChatGPT** — We have a Custom GPT with documentation already loaded ([link from support]) * **Claude** — Free at [claude.ai](https://claude.ai/), Pro recommended * **Gemini** — Free at [aistudio.google.com](https://aistudio.google.com/) ☑️ **Know your products** — What you’re searching for, what qualifies, what disqualifies --- ## Step 1: Set Up Your AI Project _~10 minutes, one-time setup_ ### Option A: ChatGPT (Easiest) We have a **Custom GPT** with all documentation pre-loaded. 1. Get the link from support 2. Open the Custom GPT 3. Skip to Step 2 — you’re ready to go **Model recommendation:** Use GPT-5.2 or 5.1 with **Thinking** or **Pro** mode enabled for complex prompts. --- ### Option B: Claude **Create the Project:** 1. Go to [claude.ai](https://claude.ai/) 2. Click **“Projects”** in the left sidebar 3. Click **“Create Project”** 4. Name it: `SKU Manager - Prompt Builder` **Add Knowledge Files:** 1. Get the SKU Manager Project Package (link from support or Google Drive) 2. Click **“Add to Project Knowledge”** (or the + icon) 3. Upload ALL files from the package 📸 _[Screenshot: Claude project with knowledge files uploaded]_ **Add Project Instructions:** 1. Find the `Project-Instructions` file in the package 2. Copy the entire contents 3. In Claude, click **gear icon** (Settings) → **Project Instructions** 4. Paste and save **Model recommendation:** Use **Claude 4.5 Sonnet** with **Extended Thinking** enabled when building prompts. --- ### Option C: Gemini 1. Go to [aistudio.google.com](https://aistudio.google.com/) 2. Create a new project 3. Upload all files from the SKU Manager Project Package 4. Add the project instructions to the **Instructions** field **Model recommendation:** Use **Gemini Pro** (thinks longer) for advanced prompt building. --- ### What’s in the Knowledge Files? The files teach your AI assistant about: * **How prompts and display templates connect** — field names, JSON structure, what needs to match * **Fields to Send** — what data comes from eBay listings * **Custom AI Columns** — optional grid columns for quick decisions * **Working examples** — reference implementations to learn from * **The Feedback/Trace system** — how to take your feedback and make targeted fixes This means your AI assistant can: * Create new prompts from scratch * Build matching display templates * Come back later and make modifications based on your feedback * Know whether a change needs the prompt, the template, or both _Note: These files may be updated over time as features evolve._ --- ## Step 2: Generate Your First Prompt & Template _~30-60 minutes_ ### Start the Conversation In your AI project, start a **new chat**. The goal is to describe what you’re looking for in natural language — like you’re explaining your buying process to someone. **Start simple:** ``` I want to create a new prompt and display template. I search for [SEARCH TERM] and see [TYPES OF PRODUCTS] in the results. When I'm going through listings, here's what I'm looking for: - [What makes something a yes] - [What makes something a no] - [Things I can only tell from pictures] - [Things I need to read in the description] ``` ### Describe Your Mental Process Think about how YOU decide whether a listing qualifies. Tell the AI assistant: * **Product identification** — “I’m looking for [specific brand/model], not [other things that show up]” * **Condition requirements** — “It needs to be [sealed/working/no damage]” * **Visual inspection** — “From pictures, I’m checking for [cracks/missing parts/box damage]” * **Description details** — “In the description, I look for [keywords/red flags]” * **Different tiers** — “I’ll pay $X for [condition A] but only $Y for [condition B]” ### Example Conversation **For Test Strips:** ``` I want to create a prompt and display template. I search for "OneTouch Ultra test strips" and I'm trying to buy sealed, undamaged boxes with good expiration dates. When I look at listings, I check: - Is it actually OneTouch Ultra (not Verio or other models)? - Is the box sealed? Any tears or damage I can see in photos? - What's the expiration date? I need at least 6 months out. - How many boxes are they selling? I want the per-box price. From pictures, I'm looking at box condition and trying to read expiration dates if they're visible. Some sellers don't show everything clearly. Let's start simple — just help me identify the right products and flag any obvious damage. ``` ### What Your AI Assistant Will Do Based on your description, it will: 1. Ask clarifying questions if needed 2. Generate a **prompt** with proper JSON structure 3. Generate a matching **display template** 4. Make sure field names align between both **Your initial goal isn’t perfection** — it’s getting the full round-trip working. You can refine it after you see real results. --- ## Step 3: Load Into SKU Manager _~5 minutes_ ### Save the Prompt 1. Copy the prompt your AI assistant generated 2. In uBuyFirst, go to **Data tab > Get Data** (External Data panel) 3. Click **“AI Prompt”** tab 4. Click **“Add”** to create new profile 5. Name it clearly: `[Product]_Prompt_v1` 6. Paste the prompt 7. Click **Save** 📸 _[Screenshot: AI Prompt tab with prompt pasted]_ ### Save the Display Template 1. Copy the HTML template 2. Go to **“AI Display Template”** tab 3. Click **“Add”** to create new profile 4. Name it: `[Product]_Template_v1` 5. Paste the template 6. Check the preview — you should see the layout structure 7. Click **Save** 📸 _[Screenshot: AI Display Template tab with preview]_ --- ## Step 4: Create Your Filter (in uBuyFirst) _~5 minutes_ **This is what triggers your AI to run.** You set this up in uBuyFirst, not in your AI project. 1. Close External Data panel 2. Go to **Home > Filters** 3. Click **“New Filter”** 4. Configure: * **Type**: Select **AI** ✅ * **Name**: Something clear like `AI - [Product] Analysis` * **AI Prompt**: Select your prompt profile * **AI Display Template**: Select your template profile * **AI Fields to Send**: Select your fields profile * **Condition**: What triggers it * For all results: `Title is not blank` * For specific items: `Alias contains "[keyword]"` or `Title contains "[model]"` 5. **Type**: Select **AI** ✅ 6. **Name**: Something clear like `AI - [Product] Analysis` 7. **AI Prompt**: Select your prompt profile 8. **AI Display Template**: Select your template profile 9. **AI Fields to Send**: Select your fields profile 10. **Condition**: What triggers it * For all results: `Title is not blank` * For specific items: `Alias contains "[keyword]"` or `Title contains "[model]"` 11. For all results: `Title is not blank` 12. For specific items: `Alias contains "[keyword]"` or `Title contains "[model]"` 13. Make sure **Enabled** is checked ✅ 14. Click **Save** 📸 _[Screenshot: Filter configuration with AI type selected]_ --- ## Step 5: Test It1. Return to **Data > Get Data** (External Data panel) 2. Click **“Stop Script”** then **“Start Script”** to refresh 3. Run a search that matches your filter condition 4. Click on a listing **Working?** You should see your formatted analysis in the External Panel. **Not Working?** Check: * Filter enabled? * Filter condition matches your search? * Field names in template match prompt output? (case-sensitive!) * API key configured and billing active? --- ## Improving Your Prompt Over Time ### When AI Gets Something Wrong Use the **Feedback/Trace System** in SKU Manager: 1. **In the External Panel**, click **Submit Feedback** on the problem listing 2. A popup appears — **describe the issue in natural language** * “This should have been flagged as damaged” * “It missed that this is a lot of 3” * “The model identification was wrong — it’s actually an X not a Y” 3. “This should have been flagged as damaged” 4. “It missed that this is a lot of 3” 5. “The model identification was wrong — it’s actually an X not a Y” 6. Click Submit — this saves everything: * Your prompt and settings * The listing data and pictures * All field data that was sent * Your feedback notes 7. Your prompt and settings 8. The listing data and pictures 9. All field data that was sent 10. Your feedback notes 11. **Go back to your AI project** and start a new chat:``` I have feedback on my [PRODUCT] prompt. The AI got something wrong. [Paste the contents of the feedback/trace files] Here's what happened: [your notes] ```1. Your AI assistant will: * Diagnose what went wrong * Show you a BEFORE/AFTER comparison * Output an updated prompt or template (or both) 2. Diagnose what went wrong 3. Show you a BEFORE/AFTER comparison 4. Output an updated prompt or template (or both) 5. **Apply the fix** in SKU Manager and re-test with the same listing ### Adding Features Later Once basics work, go back to your AI project: ``` My [PRODUCT] prompt is working. Now I want to add: - [New feature, e.g., price calculations] - [New feature, e.g., custom columns for the grid] ``` Your AI assistant knows what changes need the prompt, the template, or both. --- ## Quick Reference | What You Want | Where to Go | | --- | --- | | Create/edit prompt | Your AI project (ChatGPT/Claude/Gemini) | | Load prompt into app | SKU Manager > AI Prompt tab | | Load template into app | SKU Manager > AI Display Template tab | | Set when AI runs | uBuyFirst > Home > Filters | | Report issues/feedback | External Panel > Submit Feedback button | --- ## Adding Custom Columns (Optional) Custom columns let you see key AI decisions **in your results grid** — without opening the External Panel for each listing. ### Why Use Both? | Where It Shows | What It’s For | | --- | --- | | Display Template | Detailed analysis in External Panel — full reasoning, all the data | | Custom Columns | Quick decisions in results grid — “Can Return: Yes”, “Buy: No” | Many users include the same key values in **both** places: * Columns for quick scanning and filtering * Template for the full picture when you click in ### How to Add Columns 1. **In uBuyFirst**, go to **AI Columns** tab and create your column profile 2. Define the exact column names (e.g., `Can Return`, `Original Owner`, `Buy Decision`) 3. **Go back to your AI project** and say:``` I want to add custom columns to my prompt. Here are the exact column names I created: - Can Return (values: Yes / No) - Original Owner (values: Yes / No) - Buy Decision (values: Buy / Research / Pass) Update my prompt to output these columns. ``` **Important:** You must tell your AI assistant the **exact column names** — it can’t read them from a screenshot. The names in your prompt must match exactly what you created in uBuyFirst. ### Columns Enable Additional Filters Once you have AI columns, you can create **additional filter actions** based on the results: * Highlight rows where `AI Can Return = Yes` * Remove/hide rows where `AI Buy Decision = Pass` * Color-code based on confidence levels This lets AI do the first pass so you focus on what matters. --- ## Include Reasoning in Your Prompts When building prompts, ask your AI assistant to include **reasoning and source attribution** in the output. These sections explain back — in natural language — how and why the AI arrived at its analysis. This helps you: * Understand WHY the AI made a specific decision * Identify which field values may have caused an unexpected analysis or conflict * Debug issues faster when you submit feedback traces * See exactly which fields and photos informed each conclusion ### What Good Output Looks Like **Detailed Reasoning** (natural language explanation): ``` The phone is identified as an iPhone 15, 256GB, Factory Unlocked based on the title, description, and model fields. Return eligibility is 'Yes' as the seller has a 'Returns accepted / 30 day' policy and the device is fully functional, despite cosmetic flaws which are disclosed. The 'Original Owner' status is 'No' — the seller is a registered business with high feedback (2098) and professional product photos, strongly indicating a reseller rather than an individual. ``` **Source Attribution** (field-by-field citations): ``` Model, Storage, Carrier: Title, Model field, Description, Network field. Defects: Description, Photos 2-4. Can Return: Returns field, Description. Original Owner: SellerName, SellerBusiness, FeedbackScore, Photos. ``` When you submit feedback traces, this information helps you (and your AI assistant) pinpoint exactly what data caused the issue and where to fix it in the prompt. **Ask your prompt builder:** ``` Include a detailed reasoning section that explains in natural language how you arrived at each major conclusion. Also include source attribution showing which fields and photos informed each decision. ``` --- ## Tips for Success1. **Start simple** — Get basic identification working before adding price calculations or columns 2. **Describe your thinking** — The more you explain your mental process, the better the prompt 3. **Pictures matter** — Tell the AI what you look for in photos (damage, labels, condition) 4. **Test with real listings** — See how it performs before tweaking 5. **Use feedback traces** — Don’t guess at what went wrong; capture the actual data 6. **Include reasoning** — Makes debugging much easier when something’s off 7. **Version your work** — Save as v1, v2, etc. before making big changes --- **Questions?** Live Support at [ubuyfirst.com](https://ubuyfirst.com/) or email su*****@*******st.com _For the full AI Columns setup walkthrough, see the complete SKU Manager Setup Guide._