Amazon Auto Router

Amazon Auto Router

Automatically select the relevant specialist skills before answering or acting on an Amazon task.

Required workflow

1. Classify the request using the routing table below.
2. Select the minimum set of specialist skills that fully covers the task.
3. Read every selected SKILL.md completely before taking task actions.
4. If the task includes Feishu/Lark data, also select the matching lark-* skill for the file type.
5. Separate verified facts, model assumptions, estimates, and missing inputs.
6. Use current primary sources for fees, policies, laws, program rules, or other time-sensitive claims.
7. Complete the requested analysis or change, verify outputs, and state exact document/file locations.

Do not load every Amazon skill by default. Prefer one primary skill and add supporting skills only when they materially change the result. When amazon-listing-development-workflow or amazon-negative-keyword-batch-review is primary, let that workflow select its own specialists within the same run; do not start a competing full workflow.

Routing table

Mandatory combinations

• A new product link or material restructuring of an existing listing uses amazon-listing-regulatory-review and its publication-decision gate before publish-ready copy. For a limited edit to an already selling link, check the changed fields; invoke the full review if the user requests it or the change raises a material policy, product-function, safety, fitment, certification, or claim risk. Do not label a focused check as a full regulatory review.
• Every profit decision must include amazon-profit-analyzer; add amazon-fba-calculator when FBA fees or break-even price are involved.
• PPC work that includes search-term waste or negative targeting uses amazon-negative-keywords unless the MT002 manual batch workflow is primary; that workflow invokes the specialist only for its stated conditions and owns its review-field output.
• Every product opportunity estimate based on BSR must distinguish sales estimation from profit estimation. BSR alone cannot establish profit.
• Every Feishu spreadsheet task must use lark-sheets; every Feishu cloud document task must use lark-doc; file and folder organization must use lark-drive.

Evidence and calculation rules

• Never fabricate Seller Central, competitor, fee, sales, conversion, or inventory data.
• Treat list price, strike-through price, coupon price, and realized selling price as different values.
• For advertising economics, state whether the percentage is ACoS or TACoS. Use TACoS for total product profitability unless the user explicitly requests advertising-attributed profit.
• For returns, do not multiply price by return rate without explaining the loss model. Prefer return rate × net loss per returned order.
• For competitor profit, require or estimate price, referral fee, FBA fee, landed-cost range, ad-rate range, and return provision. Present a range, not false precision.
• Mark every temporary zero-cost input that can understate profit costs, such as storage, inbound placement, coupon fees, removal/disposal, and overhead.
• Quote formulas or show a calculation table when a recommendation depends on break-even, target margin, or restock economics.

Output standard

Lead with the decision. Then provide the supporting calculations, assumptions, risks, missing data, and the next action. When creating or updating a file, include the exact normalized title, folder path, and clickable URL or absolute local path.