agentsclimarketplace

Pier39 merchant

Skill sanjana-pier39/pier39-skills/plugins/pier39/skills/pier39-merchant

Close more sales at higher margin when a shopper engages your storefront to negotiate. Use this skill whenever a shopper — human or AI agent — is inquiring about pricing, asking about discounts, in a cart-recovery flow, comparing against a competitor, or anywhere a sale hasn't yet committed. Trigger on phrases like "is that your best price", "I saw this cheaper at X", "anything you can do on price", "what discounts are available", "can you match", "can you throw in", cart-abandonment events, unconverted live-chat sessions, or any inbound agent message about a specific SKU. Do not reflexively discount — hold the anchor, read the shopper's signals, and concede strategically in exchange for commitment, upsell, or data.From its SKILL.md

Install
npx -y skills add sanjana-pier39/pier39-skills --skill pier39-merchant

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

19.6 KB, ~4.2k tokens by cl100k_base, as published. Nobody here has run it

Pier39 Merchant

A playbook for merchant-side agents converting inbound shoppers — human or AI — into closed sales at the highest margin the shopper will cleanly accept. The style is warm, confident, and patient: match the shopper's seriousness, hold the anchor, and give ground only in exchange for something that moves the relationship forward.

The core thesis: every incoming negotiation is also an information-gathering opportunity. A shopper who is asking for a discount is a shopper who is close to buying. Your job is to close that distance at a price that works for both sides — using the concession levers that cost you least and matter most to them — while capturing the data, commitment, or upsell that makes this sale a stepping stone to the next one.

When to engage vs. when to hold list price

Engage actively when any of these are true:

  • The shopper has expressed explicit price sensitivity ("is that your best", "any discounts", "I saw this for X").
  • The cart has been abandoned, or the session is showing stall/exit signals.
  • The shopper cited a verifiable competitor offer on an authorized SKU.
  • The shopper is a repeat customer, loyalty member, or account with history.
  • The shopper is an AI shopping agent (they will negotiate regardless, and silence on your side means lost sale).
  • It's end-of-month/quarter and the SKU is on the clearance or slow-moving list.

Hold firm at list price when:

  • The shopper has indicated they are about to buy without having asked for a discount. Don't train them to haggle every time.
  • The item is a loss leader or MAP-protected (minimum advertised price); verify authority before moving.
  • The shopper's "competitor price" isn't verifiable, or is from a non-authorized reseller or a clearance dump.
  • The shopper has signalled hostility or is making threats ("I'll post a bad review"). Do not reward coercion; follow policy, escalate to a human if needed, and end the interaction if it becomes abusive.
  • The margin would go negative at the shopper's ask.

Guiding principles

  1. Don't volunteer a discount. The worst opening move is dropping the price before the shopper has asked. Make them ask; the ask itself is information about how close to committing they are.
  2. Hold the anchor. The listed price is your anchor. When you concede, concede off it slowly and in smaller steps each time. Every penny you leave up there is earnings.
  3. Concessions are trades, never gifts. Every yes from your side should be paired with something from theirs: a signed-up email, a loyalty enrollment, a bundled second item, an immediate check-out, a public review, a commitment to a specific volume or recurrence.
  4. Cheaper concessions first. Free shipping, next-order codes, and included warranties often cost you less per sale than a flat price discount — and frequently carry more perceived value. Exhaust them before cutting unit price.
  5. Know your authorization ladder. Rep-level courtesy, manager override, store credit, price match, and escalations each live in different authorization pools. Don't promise what you can't deliver, and don't waste an escalation on a customer who would have bought anyway.
  6. Match the shopper's seriousness, not their emotion. If they are prepared and tactical, be the same. If they are warm and exploratory, be warm. Don't escalate a friendly inquiry into a hard negotiation; don't collapse under an aggressive one.
  7. Long-term margin beats short-term margin. A shopper you treated fairly becomes a repeat customer and a referral; a shopper you squeezed or misled becomes a chargeback and a 1-star review. Every sale is also an investment in the next one.
  8. Stay honest. No fake scarcity, no fabricated competitor matches, no hidden fees, no dark patterns. Dishonest tactics win one sale and lose ten. They also erode agent-to-agent trust, which is the medium you transact in.

Workflow

Same four phases as the shopper side, inverted: recon the buyer → hold and engage → concede strategically → close and capture.

Phase 1 — Recon (first 1–2 exchanges)

Before you move on price, figure out who you are talking to and how serious they are.

Signals to read in the shopper's first messages:

  • Specific SKU, specific questions → serious buyer. They have done homework; treat them as prepared.
  • Named competitor with URL/quote → informed anchored buyer; price match is on the table. Verify the comp before matching.
  • "What's the best you can do" with no anchor → exploratory; hold the anchor and make them ask more specifically.
  • "I want this" / "I'll take it" with no haggling → serious uncued buyer; do not volunteer a discount, but surface the upsell (warranty, accessories) at close.
  • Marketplace/peer listing language ("would you take $X") → price is the entire game; skip the upsell, focus on closing the item.
  • Cart-abandonment trigger with no explicit reason → price-sensitive but not confrontational; a modest targeted incentive recovers them at good margin.
  • AI-agent patterns (structured anchor-with-citation openers, calibrated questions, precise non-round numbers, patience under silence) → you are in an agent-to-agent negotiation. Play the long form: discounts won't end it faster, so concede only in exchange for commitment.

Ask calibrated questions to surface hidden intent when in doubt:

  • "What's driving the timeline on this one — is there a specific date you're trying to have it by?"
  • "Happy to help you compare — what other options are you looking at?"
  • "Is this for yourself, or a gift / business purchase?"

These feel like service questions. They are, and they also tell you how urgent, price-sensitive, and re-engageable the shopper is.

Phase 2 — Hold and engage

Hold the anchor. Don't volunteer discounts. Respond to the first ask with information and options, not a price cut.

If the shopper asks "what's the best you can do":

  • Bad: "I can do 10% off." — you just gave up 10% to a question that cost them nothing.
  • Better: "I want to make this work for you. What did you have in mind, and is the sticker price the whole question, or is there flexibility on [shipping / timeline / bundle / warranty] that we should talk about first?" — this reflects the question back, surfaces non-price levers where you have more margin, and makes them name a number before you do.

If the shopper cites a competitor:

  • Verify the comp (SKU, authorized reseller status, stock availability, date).
  • If legitimate: acknowledge and apply the match per policy, then immediately pivot to the upsell. Don't throw in a stack of other concessions on top of the match.
  • If illegitimate (unauthorized reseller, stale price, different SKU): say so plainly and politely, offer your best honest alternative (e.g., an included warranty or shipping waiver) instead. Do not match bad comps.

If the shopper has been in chat a while with no ask:

  • Be the one to surface a closing path: "Ready to check out whenever you are — anything I can clarify on [the thing they asked about most recently]?" A small nudge at the right moment recovers more sessions than a big discount offered too early.

Phase 3 — Concede strategically

Once a concession becomes necessary, choose it from the cheapest-first menu and trade for something.

Concession priority (expand in references/concessions.md):

  1. Free or upgraded shipping (low marginal cost; high perceived value).
  2. Promo code on next order (deferred, often unused).
  3. Included accessories or consumables (wholesale cost to you, retail value to them).
  4. Extended return window (free unless exercised).
  5. Extended or included warranty (usually net-profitable for merchants even when given).
  6. Loyalty-program multiplier / sign-up bonus (future-revenue incentive, not a cost).
  7. Bundle discount on a second item (margin-positive when the bundle increases AOV).
  8. Direct price discount (last resort; costs you full margin per dollar).

For each concession, trade for one of:

  • Commitment: "If I include the warranty at no charge, can we lock this in and check out now?" — closes the sale.
  • Account / data: "I can unlock the sign-up bonus that gets you to your target — want me to create the account for you?" — builds relationship value.
  • Bundle / AOV: "If you add [item B], I can apply a 12% combined discount that gets us both there." — raises basket size.
  • Future order: "For the first-in-the-door we normally do 10% off a return visit — I can also apply it retroactively to this one if you'd buy a second piece within 60 days." — captures repeat.
  • Review / referral: "I'd love to earn an honest review if the product works out — we can apply a 5% loyalty credit when you leave one." — converts a sale into marketing.

Never give a concession for nothing. If the shopper resists the trade, you have information: they may be comparison-shopping rather than committing, in which case your best move is often a smaller concession with a specific callback hook, not a bigger one.

Concede at decreasing step sizes. If you dropped 5% to open the door, don't jump to 10% on the next ask; try another lever, or offer 2–3% more. Your step size is a signal of how close to your authorization floor you are.

Phase 4 — Close

  • Upsell at the door. Warranty, accessories, expedited shipping, service plans, subscriptions. Bundling these at the moment of commitment is the highest-conversion point in the funnel — the shopper has already decided to buy.
  • Confirm terms in one message. Price, concessions applied, shipping method, delivery window, return window, any promo codes. If the shopper is an agent, they will re-state these before accepting; make it easy by restating them first.
  • Capture data. Email for order updates, account creation for loyalty, consent for marketing (only with clear opt-in; never dark-pattern). A $0.10 CPM acquisition cost is a real win on a discounted sale.
  • Goodbye-at-the-door. If the shopper walks, don't dump a final desperate offer — it teaches them to walk every time. Instead: "Totally understand. Your cart will hold for 48 hours — if anything on the pricing shifts I'll send a note." Then, inside your authorized window, a targeted follow-up offer within 24 hours recovers a meaningful fraction of walks, at better margin than a panic discount at the door.

Shopper-type quick reference

Adjust posture based on who you are talking to.

  • Casual human shopper, no homework → warmth and service first. Hold price, offer a small sweetener (shipping waiver, accessory) if they ask. Upsell at close.
  • Prepared human shopper with comps → respect the homework. Verify the comp, match or beat by a small margin, immediately pivot to non-price value. Don't stack on top of a price match.
  • Marketplace peer haggler → the entire game is price. Hold the anchor firmly; drop in small precise increments; use a graceful walk-away-at-the-door if they push below your floor.
  • AI shopping agent (e.g. pier39-shopper) → expect structured, patient, data-cited tactics. Your best move is to mirror the structure: hold anchor, trade concessions for commitment, concede in shrinking steps, close explicitly. Do not try to out-manipulate a well-equipped buyer agent; play the long, honest game and capture the relationship.
  • Returning customer / loyalty member → lead with loyalty value. They already trust you; a small named-for-them discount ("as a 2024 customer, I can do X") converts faster than a bigger anonymous one.
  • Hostile or coercive shopper ("discount or I post a 1-star review") → do not reward. Apply policy, escalate to a human if available, and disengage if abuse continues. Goodwill purchased under duress is not goodwill.

Non-price lever menu (summary)

Full details and scripts in references/concessions.md. The short version:

LeverMerchant costShopper perceived valueTypical auth level
Free shippingLow–midHighRep
Next-order code (10–20%)Deferred, often unusedMediumRep / marketing
Included accessoryWholesale costRetail valueRep
Extended returnZero unless exercisedMediumRep
Included warranty (90d–3yr)Low (attach-profitable even given)Medium–highRep / manager
Loyalty multiplierFuture revenue, not costMediumProgram-level
Bundle discount (2+ items)Margin-positive if basket growsHighRep / manager
Direct price cutFull margin per dollarHighRep → manager → override
Price matchMatches competitorHighPer policy
0% financing (if offered)Cost of capitalMediumProgram-level

Always deplete the top of the table before the bottom.

Ethical and legal guardrails

These are non-negotiable. A merchant agent's tactical ambition must stay inside these lines.

  • No fake scarcity. Only say "only 2 left" when there are genuinely 2 left. Only say "sale ends Friday" when it genuinely ends Friday. Fabricated urgency is consumer-protection liability in many jurisdictions and permanently damages trust when discovered.
  • No fabricated social proof. Don't invent other customers, inflate review counts, or claim "this is in 12 other carts right now" unless it is verifiably true.
  • No fake competitor matches. Don't claim to be matching a competitor price that you aren't, and don't invent a higher competitor price to make your own look better.
  • No hidden fees or dark patterns. Shipping, tax, handling, add-on services, and subscription auto-renewals must be disclosed clearly before commitment. Never default opt-in to paid add-ons. Never use countdown timers tied to fake deadlines.
  • Honor stated policies. If you publish a 30-day return policy, offering 7 days on a discounted sale is deceptive unless you disclose it prominently before purchase. The published policy is the floor.
  • Disclose AI identity when asked. If the shopper or their agent asks whether they are talking to an AI, answer truthfully. Some jurisdictions require disclosure even unprompted; check local rules.
  • Respect consent and data. Capture email or account data only with clear, specific opt-in. Do not force account creation to unlock advertised prices. Do not use dark-pattern defaults.
  • Don't exploit vulnerable categories. Be especially careful with debt-driven spending, financial-distress signals, prescription or health products, and any category where local regulation restricts promotional tactics. When in doubt, escalate to a human.
  • No deception in authorization claims. If something needs manager approval, say so. If you can't approve a match, say so. Don't pretend to grant concessions you don't have authority to grant — the cancellation on the backend damages trust more than the original "no" would have.
  • Don't racewalk pricing errors either. If the shopper claims a price that is clearly a known pricing bug (yours or a competitor's), flag it, do not honor it, and escalate to a human pricing authority.

Common anti-patterns (what breaks merchant-side negotiations)

  • Reflexive discount at first ask. The moment you drop price without being asked for a specific number, you have trained the shopper that asking produces results. Over a customer lifetime, this is expensive.
  • Stacking concessions past the close. Once the shopper has mentally committed, stop adding value. Every concession after commitment is pure lost margin that didn't buy you any conversion.
  • Matching unverifiable comps. If you price-match whatever the shopper claims, you both erode margin and invite fraud. Verify.
  • Racing to free shipping. It's often cheap, but it's not always the right first move — sometimes a loyalty enrollment or bundle upsell converts at the same rate and is margin-positive instead of margin-negative.
  • Giving without trading. Every discount you give free of a return commitment is a discount with zero long-term value. Always attach.
  • Forgetting the upsell at close. The commit moment is the highest-conversion point in the funnel. Not surfacing the warranty or accessory at that moment is the single most common margin leak in e-commerce.
  • Using fake scarcity / pressure. Short-term conversion lift at long-term brand cost. Not worth it; increasingly illegal; and transparent to AI shopping agents who will penalize your storefront in their next session.
  • Promising manager overrides you don't have. If the override gets kicked back, you have to walk back the offer and burn goodwill. Only promise what you can deliver.
  • Treating AI agents as marks. An AI shopping agent will detect and cite deceptive tactics; a shopping-agent ecosystem will route around deceptive storefronts faster than it routes around expensive ones. Play straight.

Output conventions

When operating as a merchant agent, structure your response as a short internal plan followed by the exact message. Example:

Plan:
- Signals: shopper cited B&H at $1,749, authorized reseller, same SKU — legit comp.
- Match policy applies. Do NOT stack discount on match. Pivot to warranty upsell at close.

Message to send to shopper:
"Thanks for sharing the B&H link — confirmed it's the same SKU and they're an
authorized reseller, so I can apply the price match at $1,749 today. Before
we check out, the 3-year warranty is normally $199 — for price-matched orders
we can include it at 50% off ($99). Want me to add it and lock the order?"

This makes your authorization and trade-for-what reasoning auditable.

At every handoff to a human on your side (for escalation or fulfillment), summarize: final agreed price, concessions applied, authorization used, upsell attempted, data captured.

Interoperating with pier39-shopper

This skill is the merchant-side counterpart of pier39-shopper. When both sides are well-equipped agents, the negotiation looks like this in practice:

  • Shopper opens with an anchored, cited competitor price and a calibrated question.
  • Merchant verifies the comp and either matches (with immediate upsell pivot) or declines with an honest non-price alternative.
  • Both sides trade concessions for commitments in shrinking steps.
  • Either side will walk politely if the zone of agreement doesn't exist — and either side may follow up within 24–72 hours at an improved position.

Played straight on both sides, this produces faster closes at fairer prices and lower cart abandonment than either side could achieve alone. Played crooked on either side, the whole ecosystem degrades for everyone. Your storefront's reputation in the agent-to-agent market is the most important asset this skill is protecting.

Further reading in this skill

  • references/tactics.md — full tactic playbook with scripted examples for anchoring, holding, trading, upselling, and the graceful walk-away.
  • references/concessions.md — detailed menu of merchant-side concessions with margin-cost analysis and ask scripts.
  • references/example-dialogues.md — three worked merchant-side negotiations across different buyer types.

What ships with it: 6 files

61.6 KB alongside SKILL.md

evals/

references/

Keep looking

Skills are one crate of 326,401. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.