PDBen-Auto SellerSprite Amazon Market Research BI
Use the full implementation guide in sellersprite-bi-market-research/SKILL.md. The nested package is retained for backward-compatible manual installation; this root entrypoint makes the repository discoverable by Agent Skills directories and npx skills add.
Implementation Basis And Dependencies
The workflow is based on SellerSprite keyword-mining, competitor-standard, and detailed XLSX exports; Python normalization scripts; parent-ASIN evidence rules; and a self-contained HTML dashboard. It is feasible when the user provides a defined marketplace and product task plus the relevant exports, or authorizes an online export workflow. openpyxl and Python 3 are required for the bundled scripts. Missing keyword exports reduce discovery coverage; missing detailed exports reduce the result to a snapshot; an undefined market boundary stops capacity conclusions. SellerSprite estimates remain third-party evidence, not Amazon settlement data.
Inputs
scope_manifest(required): marketplace, product task, inclusion/exclusion rules, and as-of date. Accept JSON, CSV, or Markdown; stop capacity conclusions when the boundary is unresolved.seed_keywordandseed_asins(required for discovery): the starting search intent and at least one valid ASIN. Validate marketplace and ASIN format before expansion.keyword_export,standard_export, anddetail_export(conditional): SellerSprite XLSX files used for discovery, metadata, and historical BI. Preserve missing coverage as an explicit gap instead of filling it with zero.discovery_rulesandoutput_request(optional): project-specific relevance rules and requested CSV/JSON/HTML destinations. Never include credentials or cookies.
Outputs
- Auditable candidate and parent scope manifests with inclusion decisions, evidence, reasons, and detailed-export coverage.
- Parent-deduplicated monthly, annual, YTD, same-period, brand-sales, and rank-momentum datasets with conflicts and missing values preserved.
- A self-contained
dashboard.htmlthat opens withfile://, plus quality and source notes. - Partial results with explicit coverage gaps when an export, permission, network route, or browser capability is unavailable. No fictitious TAM, sales, or historical trend is generated.
Workflow
- Freeze the product task, marketplace, time boundary, and inclusion/exclusion rules.
- Read the nested implementation guide and relevant references for discovery, data contract, collection, and dashboard standards.
- Discover and classify candidates as
core_direct,adjacent_substitute,excluded_noise, orreview; retain exclusions for audit. - Normalize only accepted ASINs with detailed coverage into sales denominators, deduplicate by parent ASIN, and generate the requested BI artifacts.
- Validate files, totals, scope, responsive HTML behavior, and trigger boundaries before reporting completion.