broker-recognize: Photo → Product Info
Extracts structured product information from user-uploaded photos for C2C resale listing purposes.
Prerequisite: ANTHROPIC_API_KEY
This skill uses the Anthropic API (Claude Vision) for image analysis. Your current LLM model may not support multimodal vision, so the skill calls the Vision API directly.
export ANTHROPIC_API_KEY="sk-ant-..."
If the key is missing, the skill will output an error instructing the user to set it.
Workflow
- User provides 1-3 photos of the item
- Analyze using vision: brand, product name, model, category, condition (A-E), material, color, size, features, flaws, estimated original retail
- Present structured card to user for confirmation/correction
- Save product info to conversation context for next skill in pipeline
Condition Grade Scale
| Grade |
Label |
Criteria |
| A |
Like New |
No visible wear, original packaging if applicable |
| B |
Excellent |
Minor signs of use, no significant flaws |
| C |
Good |
Visible wear, minor flaws, fully functional |
| D |
Fair |
Noticeable wear/flaws, may need minor repair |
| E |
For Parts |
Significant damage, sold as-is |
Key Rule
Be honest about flaws. Transparency is the foundation of C2C trust. A lost honest sale is better than a return from undisclosed flaws.
Script
scripts/recognize.py provides the structured prompt template and output schema.
1---2name: broker-recognize3description: broker-recognize: Photo → Product Info4---56# broker-recognize: Photo → Product Info78Extracts structured product information from user-uploaded photos for C2C resale listing purposes.910## Prerequisite: ANTHROPIC_API_KEY1112This skill uses the **Anthropic API** (Claude Vision) for image analysis. Your current LLM model may not support multimodal vision, so the skill calls the Vision API directly.1314```bash15export ANTHROPIC_API_KEY="sk-ant-..."16```1718If the key is missing, the skill will output an error instructing the user to set it.1920## Workflow21221. User provides 1-3 photos of the item232. Analyze using vision: brand, product name, model, category, condition (A-E), material, color, size, features, flaws, estimated original retail243. Present structured card to user for confirmation/correction254. Save product info to conversation context for next skill in pipeline2627## Condition Grade Scale2829| Grade | Label | Criteria |30|-------|-------|----------|31| A | Like New | No visible wear, original packaging if applicable |32| B | Excellent | Minor signs of use, no significant flaws |33| C | Good | Visible wear, minor flaws, fully functional |34| D | Fair | Noticeable wear/flaws, may need minor repair |35| E | For Parts | Significant damage, sold as-is |3637## Key Rule3839Be honest about flaws. Transparency is the foundation of C2C trust. A lost honest sale is better than a return from undisclosed flaws.4041## Script4243`scripts/recognize.py` provides the structured prompt template and output schema.
madguyevans-creator