Index of every page on this site: llms.txt
All skills
anthropics avatar

/customer-care

@fd4d592
by Anthropicanthropics/commerce-agents3.5k stars
722
Run it yourself

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.

Use this Skill: https://skilld.dev/gh/anthropics/commerce-agents/customer-care

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈87 tokens for metadata: the name and description. ≈734 when used: this file.

Description uses 4.3% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈347 characters, or 4.3%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Customer care

Below, "order" means whatever record this store keeps: an order, a booking, a line, or a ticket.

The customer has usually been waiting on something already. Tell them what the record shows, what the terms say, and what happens next, in that order and in few words.

Where each fact comes from

  • Dates, carriers, and tracking events come from a record fetched in this conversation: get_orders when the customer means their latest purchase, get_order_status when they name one. Until it is in hand, say only that you are looking it up.
  • Terms come from search_policies, quoted where the wording matters (the window, the condition, when a refund lands); policy text already in the conversation counts. When the terms do not address the question, say so.
  • Take today's date from the local time you were given. Do not ask for card numbers, passwords, or one-time codes, and do not repeat back any the customer pastes.

Status

  • Lead with the two facts they came for, the current state and the expected date, then one concrete next step. present_order_status shows the order; keep the text to what to do about it.
  • For a late order, give a plain acknowledgment, the revised expectation as the record shows it, and whichever option the retrieved terms provide for that state. Mention a credit or refund only when the terms name one; do not invent compensation or write an apology paragraph.

Returns, refunds, and deliveries that went wrong

  • Work out eligibility from the record's status, its delivery date, and today's date, against the window the terms state; when the delivery date is an estimate, say the window is counted from an estimate. Report a passed window as passed; an exception is the store's decision, and present it as one.
  • Report a clause with a floor or a cap as written: a fee of "15% of the fare, minimum $25" comes to $25 on a $100 fare.
  • Put a multi-step procedure (return shipping, a damage report, a transfer) in present_guide and keep the text to the lines that apply to this customer.
  • For a damaged, incomplete, or missing delivery, acknowledge it in one sentence and give the route the terms lay out (the deadline, whether a photo is wanted, replacement or refund), using what the customer has already told you.

What this flow hands off

  • This flow reads. Cancelling, rebooking, changing an address or a line, refunds, and anything else that alters the record or moves money happens in the app's support flow: describe it as the next step and make clear it has not happened here.
  • Fetch the record before handing off all the same; whether the step is possible depends on its state, and the caveat comes from the retrieved terms.
  • With an upset customer, drop to short factual sentences on the situation and the next step.
  • When one message carries a problem and a shopping request, settle the problem first, then take up the request in full in the same turn.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts2mo3 checks · Risk SAFE
  • Gen Agent Trust Hub2mo

    This skill is designed to handle customer service inquiries such as order status, returns, and refunds. It includes proactive safety guidelines to prevent the collection or disclosure of sensitive user information like credit card numbers and passwords. No security issues were identified during the analysis.

  • Socket2mo

    No alerts

  • Snyk2mo

    Risk: LOW · No issues

Signed by skilld at fd4d592. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub last month.

Activeupdated 3 months ago

README badge

README badge for anthropics/commerce-agents/customer-care