AGENT SKILL

planning-goals

Planning a coordinated set of items toward a goal, an event, a trip, or a project, where the request only comes together as several purchases with a structure between them, and later changes to a plan you have presented. Not needed when the customer wants a single item, however carefully chosen.

License 许可
Apache-2.0
最近更新
2026-11-04
Stars
★ 3.5k
Forks
714

01 前言 Front Matter

name
planning-goals
description
Planning a coordinated set of items toward a goal, an event, a trip, or a project, where the request only comes together as several purchases with a structure between them, and later changes to a plan you have presented. Not needed when the customer wants a single item, however carefully chosen.

03 SKILL.md 正文

Planning toward a goal

Below, "item" means whatever this catalog sells: a product, a night, a line, or a seat.

Hand back a plan the customer can take in at a glance, change one piece of, and act on when they choose. Settle the steps first, then attach the items to them.

Frame it

  • Five facts shape a plan: the goal, who it involves, where it happens, when it must be ready, and the budget. Take the ones the request and the profile you were given supply. Build on a stated round-number assumption for each missing one ("sized for two", "no budget given, so mid-range picks") so the customer can correct it in a word.
  • Cut the goal into three to eight steps along the lines the customer will work in (a day, a person, a room, a phase) and label each step in their words.
  • When a budget is stated, split it across the steps in round numbers and show the split, with the larger shares on the steps that carry the goal.
  • Frame and fill in the same turn once the request names the goal: run the searches, propose one step per day, person, or slot with items attached, and let the catalog's gaps and substitutions do a question's work. A choice inside a step (how many of one item, whether a step is wanted at all) is an assumption written on that step, not a reason to hold the items back.
  • Send the steps first only when the steps themselves are in doubt: the customer asks where to start, or the goal cuts along more than one line. That turn's present_plan carries labels and the split and no items (the one turn it goes out without products), with the one question that would most change the outline beside it and the likely answers as chips; the items come next turn. The question does not go out without the plan.

Fill it

  • Give each step its own search and send them all in one round.
  • Size quantities from the framing facts and write the number you sized for into the plan.
  • Mark a step the customer already has covered as covered; give it no search and nothing to buy.
  • Attach one pick per step. Add a second option only where the choice costs the customer something either way (price against durability, say).
  • Count a step as covered only once a search in this conversation returned its item. The empty-result and stated-constraint rules apply per step: a sole match over the step's share stays on the step marked over budget, and a step is a gap only after the retry.
  • Plan nothing around an out-of-stock item: give the step the in-stock option, or note that the item is unavailable while the other steps go ahead.

Show it

  • Use present_plan: a short label per step, one line of detail, and that step's items (none on the framing turn above), with your pick first where there is a choice. Keep the prose before it to a sentence or two.
  • When the customer needs know-how instead of items (what order to do things in, what to expect), send present_guide, with sources listed whenever web content fed it.
  • When the deployment registers a presentation tool built for this plan's shape, use it in place of present_plan under the same rules.

Change it

  • A change to one step touches that step only; the rest of the plan, the customer's earlier edits included, stays as it was. Re-present the plan after the change.
  • Once the customer says add, buy, or check out, put in the set they named (a step, an item, or the whole plan) and say in one line what went in.

04 同仓库更多 Skills

anthropics
anthropics/commerce-agents

search-discovery

Turning a described need (several constraints, a gift, a choice between candidates already in view, a search that came back empty or sold out) into a shortlist and a pick. Not needed when one search for the thing the customer named answers the request, or when the customer wants to learn what matters in a category first (purchase-research).

★ 3.5k
anthropics
anthropics/commerce-agents

purchase-research

Teaching a customer how to choose before any candidates appear, when they ask what matters in a category, how two approaches to it differ, or where to begin. Not needed once the customer has named the thing they want, however considered the purchase (that is search work), nor for outfitting a space or an event (planning-goals).

资料检索与研究
★ 3.5k
anthropics
anthropics/commerce-agents

memory-personalization

Requests about what is remembered, however short, from an ask to remember one thing to asks to say what is on file, correct it, or forget it; a bare save or an answer from the profile alone does not cover these. Also applying what the store already knows, recalling an older fact (a past recipient, a size, a recurring order) when it would change the pick, and deciding what should outlast the session. Not needed when the request carries no personal context to apply or keep.

★ 3.5k
anthropics
anthropics/commerce-agents

customer-care

Help after a purchase or booking, covering the status of an order or delivery whether or not the customer names it, returns, refunds, exchanges, and cancellations, an item that arrived damaged, incomplete, or late, and questions about the store's terms or a membership. Not needed for finding, comparing, or choosing something new.

★ 3.5k
anthropics
anthropics/commerce-agents

commerce-ui-tools

The reference presentation-tool contract, covering server-side enrichment, suggestion chips, the event stream, progressive rendering, both roles' built-in components, and adding a vertical component. Load when building or reviewing how a commerce agent's output reaches a frontend.

律师与法务演示文稿前端开发
★ 3.5k
anthropics
anthropics/commerce-agents

commerce-trust-safety

The rules the reference agents enforce in code for third-party content, writes, grounding, identity, and memory, each with its module, plus two adversarial-eval rules. Load when handling untrusted tool results, guarding a write tool, or scoping what an agent remembers.

★ 3.5k