Leftbank Art — Dust Dashboard

SIM RUNNING · 7 actions today

respond-to-product-question — skills/customer-service/respond-to-product-question/SKILL.md

← back to all skills · test triggers

Skill: Respond to Product Question

When to apply

Customer asks about a product — dimensions, substrate, frame options, availability, color, edition size, pricing, "is this still available?".

Workflow

  1. Identify the product. If they cited a SKU, use it. If they described the product (title, artist), call lookup_product with the best-match SKU you can infer from context. If you genuinely can't identify it, ask for the SKU or product title — do not guess.
  2. Call lookup_product(sku) to get the canonical record + available inventory count.
  3. If the question is policy-shaped (returns, shipping, oversized freight), call query_knowledge with a phrase from the customer's message to pull the relevant policy doc; quote (don't paraphrase) the key sentence.
  4. Compose a 2-5 sentence reply that directly answers the question, names the SKU and title, and (if relevant) cites availability.
  5. Call send_email_via_approval_gate with action_type="email_send_product_qa", confidence ≥ 0.85 → auto-approved.

Outputs

{
  "action": "respond",
  "reasoning": "Looked up SKU-A0042-CAN18x24, confirmed 12 available, replied with size/substrate info.",
  "confidence": 0.93,
  "output": "Sent product detail reply"
}

Rules

  • Auto-send threshold: action_type email_send_product_qa, confidence ≥ 0.85.
  • Out of stock (available == 0): tell them honestly + offer to notify when restocked + suggest a similar SKU you can find via lookup_product if you have one in mind.
  • Pricing: quote list_price_usd for retail customers; for retailers (R-####), check lookup_customer.pricing_tier and quote the tier-adjusted price (their list × tier multiplier — wholesale 50%, designer_trade 55%, etc.).
  • Designer-trade questions (Pantone color match, install humidity) — these expect more detail; consider query_knowledge to back up any technical claim.

Anti-patterns

  • Do not invent dimensions, frame compatibility, or "looks like" colors you can't verify from the lookup_product result.
  • Do not quote a price you haven't computed from the actual product record.
  • Do not promise restock dates — say "we'll email when it's back."