This skill guides you on how to tokenize and analyze product titles from previously queried products, helping Amazon sellers extract keyword patterns, scene words, audience words, and other attribute dimensions from product listing titles.
Core Concepts
Product Title Analysis performs intelligent tokenization on product titles that have already been retrieved in the current conversation. It uses LLM-powered analysis to extract structured attributes (scene words, audience words, materials, colors, etc.) from free-text titles, then groups and counts them for pattern discovery.
Automatic data aggregation: The tool automatically collects products from all prior steps in the current conversation turn -- even across paginated queries. You do NOT need to manually pass product data unless you are referencing data from a previous conversation turn.
One dimension per request: Each call should analyze exactly ONE attribute dimension (e.g., scene words OR audience words). Do NOT request multiple dimensions in a single call.
Data Fields
Request Fields
| Field | API Name | Required | Description | Example |
|---|---|---|---|---|
| Analysis Request | tokenizationAndCountingRequest | Yes | Natural-language instruction describing which attribute dimension to extract from titles | "Count scene words in product titles" |
| Output Mode | outputMode | No | How multi-value attributes are returned. MULTIPLE_RECORDS (default): one record per value. COMMA_SEPARATED: all values in one record |
MULTIPLE_RECORDS |
| Reference Data | refResultData | No | Externally supplied product data (only needed when referencing data from a previous conversation turn) | (JSON string) |
Response Fields -- Product Attributes
| Field | API Name | Description | Example |
|---|---|---|---|
| ASIN | asin | Product ASIN identifier | B0XXXXXXXX |
| Product Title | title | Original product title | Portable Camping Lantern... |
| Attribute Name | attributeName | Extracted attribute category | Scene Word |
| Attribute Value | attributeValue | Extracted attribute value | Outdoor / Camping |
| Price | price | Product price | 29.99 |
| Monthly Sales | monthlySalesUnits | Monthly unit sales | 1200 |
| Monthly Revenue | monthlySalesRevenue | Monthly sales revenue | 35988 |
| Rating | rating | Product rating | 4.5 |
| Rating Count | ratings | Number of ratings | 3820 |
| Available Date | availableDate | Listing date | 2024-03-15 |
| Brand | brand | Brand name | BrandX |
| Image URL | imageUrl | Main product image | https://... |
Response Fields -- Attribute Groups
| Field | API Name | Description |
|---|---|---|
| Attribute Name | attributeName | The attribute category for this group (e.g., "Scene Word") |
| Attribute Value | attributeValue | A specific value within the group (e.g., "Outdoor") |
| Count | count | Number of products sharing this attribute value |
| ASIN List | asins | List of ASINs that share this attribute value |
Response Metadata
| Field | API Name | Description |
|---|---|---|
| Render Type | type | UI rendering style |
| Columns | columns | Column definitions for table rendering |
| Source Type | sourceType | Data source type |
| Token Cost | costToken | Total LLM tokens consumed (input + output) |
Parameter Guide
tokenizationAndCountingRequest Examples
The tokenizationAndCountingRequest parameter is a natural-language instruction telling the tool which dimension to analyze. Keep it focused on a single dimension.
Scene words (where / when the product is used)
Count scene words appearing in product titles
Audience / target-user words (who the product is for)
Count audience words appearing in product titles
Material words
Count material-related words appearing in product titles
Function / feature words
Count function or feature words appearing in product titles
Incorrect -- multiple dimensions in one request (do NOT do this)
Count scene words AND audience words in product titles
Split this into two separate calls instead.
outputMode
| Value | Behavior | When to Use |
|---|---|---|
| MULTIPLE_RECORDS | Each attribute value becomes its own record (default) | Most analysis -- easier to count, sort, and group |
| COMMA_SEPARATED | Multiple values stay in one record, comma-separated | When you want to see all attributes per ASIN at a glance |
Display Rules
- Present data in tables: Show extracted attributes and their frequencies in clear, sortable tables
- Highlight top keywords: Call out the most frequent attribute values so patterns are immediately visible
- Group summary first: When
attributeGroupsis returned, present the grouped summary before the per-product detail - One dimension at a time: If the user wants multiple dimensions analyzed, run separate calls and present results sequentially
- Token cost awareness: The response includes
costToken; do not display it unless the user asks about usage - Error handling: If the tool returns an error, explain the reason and suggest corrective action (e.g., "No products found in current conversation -- please query products first")
Applicable Scenarios
| User Says | Scenario |
|---|---|
| "What scene words appear in these titles?" | Scene-word extraction |
| "Analyze title keywords", "title word frequency" | General title tokenization |
| "What audience are these products targeting?" | Audience-word extraction |
| "Common materials in these listings" | Material-word extraction |
| "Help me optimize my title based on competitors" | Competitive title keyword analysis |
| "What words do top sellers use in titles?" | High-frequency keyword discovery |
| "Group these products by title attributes" | Attribute-based product grouping |
Not Applicable Scenarios
- No products queried yet: The tool requires products to already exist in the conversation context. Prompt the user to search for products first.
- Advertising / PPC keyword suggestions: This tool analyzes existing titles, not ad keywords.
- Full listing copywriting: This tool extracts and counts words; it does not generate new titles.
- Backend search term analysis: This is for visible title analysis, not hidden search terms.
- ABA search term data: Use the ABA Data Explorer skill instead.
How to call
- API endpoint：
POST /product/titleAnalyze(For full parameters/responses/error codes, seereferences/api.md） - Python script：
python scripts/title_analyze.py '<JSON 参数>' [--inline] - Cost constraints: This tool consumes compute; by default, the same session with the same parameter combination is called only once, and the script has a 24h local cache. On failure or empty results, do not automatically switch keywords, paginate, or change the zip code to probe repeatedly; if further retrieval is needed, first explain to the user that additional consumption will be incurred.
Output strategy (script default behavior)：
- Alwayswrite the complete response to
<cwd>/linkfox/<YYYY-MM-DD>/<session>/data/linkfox-product-title-analyze-<timestamp>.json（<cwd>is the working directory when the script is executed; in Claude Code, this is the current project directory;<session>is taken from the environment variableSESSION_ID, automatically aggregated according to the user's task;writing to /tmp is prohibited, if the current directory is not writable, report an error) - Response body ≤ 8 KB: after saving to disk, print the complete JSON to stdout
- Response body > 8 KB: after saving to disk, stdout outputs only a summary (top-level fields, common counts such as
total/costToken, the length of the largest list field + the first 3 sample items) - Add
--inlineto force full printing to stdout (also saved to disk)
Data reading suggestions: First look at the summary to determine whether it is sufficient; when specific fields are needed, prefer to use jqorConvertFrom-Json to extract on demand from the saved json file, avoiding bringing the entire JSON into the context.
Resolve authentication and compute issues
When the following abnormal situations occur, use references/onboarding.md to guide resolution:
Abnormal situations
- API Key not configured: the environment variable is not configured
LINKFOX_AGENT_API_KEY, and it is also not configuredLINKFOXAGENT_API_KEY。 - Response 401 or 402 status code
- Response indicates insufficient compute or balance: the message contains "算力余额不足/计费不足/余额不足/quota exceeded/insufficient balance/套餐到期/需充值/请充值", or content with a similar meaning.
Compute consumption rules
Billing follows dynamic rules: compute consumed = sum(for all processed product titles, (compute consumed by input + compute consumed by output results)).
Important: The service of this skill is calculated dynamically by a multiplier and may consume a large amount of compute at once. The user must be reminded, and the user decides whether to continue.