Amazon Review Intelligence
Collect written reviews without requiring an Amazon login on the primary path, retain evidence, export Excel/JSON, and connect recurring VOC themes to product hypotheses.
PDBen-Auto / Agent Skills
Evidence-driven Amazon product research
A connected set of installable workflows for finding the right market, preserving customer evidence, reducing design/IP risk, and making a traceable product investment decision.
Market discoveryCustomer evidenceDesign/IP riskProduct validationGo / No-Go
Start with the evidence you already have
Keywords or SellerSprite exportsBuild an auditable Amazon market boundary.
ASINs, reviews, or customer pain pointsTurn written customer evidence into product hypotheses.
Product sketches or visual IP concernsPre-screen design risk and plan structurally distinct alternatives.
A complete evidence bundleValidate the handoff to a product Go/No-Go decision.
Choose a Skill
Each repository performs one coherent job. Install only the component you need, or combine the outputs in the Decision Gateway.
SellerSprite Amazon Market Research BI
Expand keywords and competitor ASINs, filter relevant listings, deduplicate parent ASINs, and create a self-contained, auditable Amazon category dashboard.
Design Patent Search And Design Around
Explain drawing-based visual scope, separate substantive and marketplace risk, and turn design-around work into supplier questions and prototype gates.
Amazon Product Decision Gateway
Validate a sanitized evidence bundle and verify an official signed stage-gate result from a separately hosted private decision engine.
What the output looks like
A synthetic 20-review case is converted into an evidence workbook and a traceable product decision report. The same delivery pattern is used across the collection: source boundary, normalized records, explicit calculations, and a portable artifact.
Worked cases
Each case states the input boundary, evidence handling, output, and what remains for human review.
Shared release standard
Every package documents inputs, outputs, dependencies, evidence limits, and stop conditions. Public examples are synthetic or redacted. Provenance is public and prompt-neutral; no hidden instruction, covert telemetry, or backdoor is part of the release design.