amazon-product-research-agent-skills

Amazon Product Research Suite

Purpose and scope

Use this Skill as the suite entry point when the product decision crosses two or more evidence domains, or when the user does not know which specialist Skill applies. Select the smallest sufficient set of Skills, preserve the marketplace and evidence date, and make missing inputs explicit before execution.

For a narrow request, route directly to the specialist and do not run the suite merely to produce a longer report. The router never invents market data, legal conclusions, supplier promises, cost assumptions, or a Go/No-Go score.

Implementation Basis and Dependencies

The router uses the Agent Skills directory contract, the routing table below, and the bundled artifact schemas in references/. It sequences public specialist workflows without copying private scoring logic into this package.

Dependency	Required	Provider and validation	Fallback or stop behavior
Compatible Agent Skills runtime	Yes	Host runtime; confirm it can load this SKILL.md	Return a manual routing plan
One or more specialist Skills	Conditional	Local installation or the public repositories named below	Give the exact install command; do not claim execution
Product evidence	Conditional	User-authorized files, URLs, exports, images, quotes, or assumptions	Return an evidence-gap plan
Network access	Conditional	Required only for live public-source collection or installation	Work from supplied evidence or stop the live step
Python 3.10+	Optional	Runs the zero-third-party-dependency installer and validators	Install manually and inspect JSON artifacts manually
Credentials or private services	Not required by router	Never request or store them for routing	Keep private execution outside the public package

Feasibility requires a defined product decision and enough scope to identify the relevant marketplace, jurisdiction, evidence date, and business constraint. Live collection is limited to public or user-authorized sources.

Inputs

Input	Requirement	Type, source, validation, and sensitivity
Product decision	Required	Plain-language question naming the decision to make; ask for one sentence if unclear
Product identifiers	Conditional	ASIN, URL, SKU, images, model, or description; preserve identifiers exactly
Scope	Conditional	Marketplace, jurisdiction, evidence date or time window; required when conclusions depend on location or recency
Evidence files	Optional	CSV, JSON, XLSX, PDF, image, URL, or supplier quote; record source and date and never silently upload sensitive data
Business constraints	Optional	Price, margin, MOQ, tooling budget, lead time, return-rate assumption, cash limit, and non-negotiable features
Authorization boundary	Conditional	State whether public-source collection, account use, external processing, or uploads are allowed

If inputs are stale, conflicting, sensitive, or incomplete, report the smallest blocking gap. Never infer credentials, legal jurisdiction, a market denominator, supplier commitments, or missing financial assumptions.

Outputs

Every run returns:

1. route-plan.json or an equivalent structured routing record conforming to references/route-plan.schema.json.
2. Selected Skill names, reasons, pinned install commands when missing, and the smallest execution order.
3. An evidence contract conforming to references/evidence-bundle.schema.json for cross-Skill handoffs.
4. Missing inputs separated into blockers and optional quality gaps.
5. Expected artifact destinations such as JSON, CSV, XLSX, HTML, search logs, or decision handoffs.
6. Explicit boundaries and partial-result status.

Success means the user can identify what runs, what evidence it needs, what artifact it produces, and what stops it. External writes, downloads, account actions, or uploads require the authorization normally required by the selected Skill; routing alone causes none of those side effects.

Routing rules

User intent	Specialist Skill	Install command when missing
Reviews, complaints, VOC, recurring pain, review collection	amazon-review-scraper	npx skills add PDBen-Auto/amazon-review-intelligence-skill --skill amazon-review-scraper
Convert evidence into differentiated mechanisms, specifications, experiments, and kill criteria	amazon-product-differentiation-rd	npx skills add PDBen-Auto/amazon-product-research-agent-skills --skill amazon-product-differentiation-rd
Visual similarity, design rights, design patent, design-around	design-patent-search-and-design-around	npx skills add PDBen-Auto/design-patent-design-around-skill --skill design-patent-search-and-design-around
Amazon category sizing, keyword expansion, ASIN discovery, SellerSprite, market boundary	sellersprite-bi-market-research	npx skills add PDBen-Auto/sellersprite-amazon-market-research-bi-skill --skill sellersprite-bi-market-research
RFQ, MOQ, tooling, lead time, supplier comparison, sample gates	amazon-supplier-feasibility	npx skills add PDBen-Auto/amazon-product-research-agent-skills --skill amazon-supplier-feasibility
Contribution margin, break-even ACOS, returns sensitivity, inventory cash	amazon-unit-economics-cashflow	npx skills add PDBen-Auto/amazon-product-research-agent-skills --skill amazon-unit-economics-cashflow
Stage gate or formal Go/No-Go handoff	amazon-product-decision-gateway	npx skills add PDBen-Auto/amazon-product-decision-suite --skill amazon-product-decision-gateway

For a cross-functional launch decision, normally sequence market -> customer evidence -> differentiation -> design/IP -> supplier -> unit economics -> decision gateway. Skip any module that cannot change the decision.

Workflow

1. Restate the decision, marketplace, time window, and deadline in one sentence.
2. Select the smallest set of relevant Skills from the routing table above.
3. Check installation and required inputs before promising execution.
4. Create the route plan and evidence contract.
5. Run specialist Skills only when their data access and external actions are authorized.
6. Preserve source meaning and label each claim as observed, calculated, modeled, inferred, or assumed.
7. Validate cross-Skill JSON against the bundled schemas when those artifacts are produced. In the repository checkout, scripts/validate_artifact.py provides a zero-dependency validator.
8. Stop at a hard blocker, preserve partial results, and state the recovery action.
9. Verify every promised artifact exists before reporting completion.

Example

Use $amazon-product-research-suite to evaluate a magnetic car phone mount for Amazon US.
I need the direct market, recurring complaints, a differentiated specification, supplier
questions, unit economics, and a recommendation on whether to prototype.

Route to market -> review -> differentiation -> supplier -> unit economics -> decision gateway. Add design-patent pre-screening when product images or a design direction exist. Required before execution: identifiers or seed terms, evidence date, target price, cost assumptions, cash limit, and authorization for any live collection.

Error handling and stopping conditions

• Missing specialist: return its pinned catalog install command.
• Unsupported marketplace or jurisdiction: narrow scope or stop the affected conclusion.
• CAPTCHA, sign-in warning, account challenge, or access restriction: stop collection and preserve partial data.
• Stale or incomplete export: downgrade coverage and label the limitation.
• Missing cost, MOQ, return-rate, or advertising assumptions: do not issue a financial recommendation.
• Private engine unavailable: create a sanitized local handoff; do not guess its result.
• Sensitive data detected: redact, obtain authorization, or stop before external processing.

Validation and definition of done

• Narrow review, differentiation, patent, market, supplier, finance, and decision requests each select one correct specialist.
• Cross-functional requests produce an ordered multi-Skill plan.
• Generic ecommerce copywriting, listing writing, and unrelated data analysis do not trigger this router.
• Missing specialists produce pinned install instructions rather than fabricated execution.
• python scripts/validate_catalog.py, python scripts/validate_artifact.py, and the test suite pass.

This site is open source. Improve this page.