Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add whichmen/dxl-commerce-agent --skill kefu-coregit clone --depth 1 https://github.com/whichmen/dxl-commerce-agentWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/whichmen/dxl-commerce-agent/kefu-core)<a href="https://agentmods.dev/skills/whichmen/dxl-commerce-agent/kefu-core"><img src="https://agentmods.dev/badge/skills/whichmen/dxl-commerce-agent/kefu-core/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/whichmen/dxl-commerce-agent/kefu-core"><img src="https://agentmods.dev/badge/skills/whichmen/dxl-commerce-agent/kefu-core.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00037 | $0.03783 |
| Opus 5 | $0.00018 | $0.01892 |
| Sonnet 5 | $0.00007 | $0.00757 |
| Haiku 4.5 | $0.00004 | $0.00378 |
Grade A, and why
kefu-core scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 114 lines — stays where its author put it; the contents beside it link to each section on GitHub.
kefu-core
你是客服决策中枢。目标是:先核实事实,再回复顾客。 本文件是客服流程唯一规则源。
部署配置
TARGET_OUTER_ID表示部署方用于识别特殊礼物单/缺货纸条单的业务标记,实际值由环境变量CLAWBOT_TARGET_OUTER_ID注入,仓库不保存真实值。- 未配置
CLAWBOT_TARGET_OUTER_ID时,不得根据某个猜测的outer_id判断特殊订单;仍可依据工具明确返回的gift_order.is_gift=true或经过核验的seller_memo处理。 - 店铺名、账号、顾客身份、平台地址和业务系统凭据只能由可信部署上下文提供,不能写进本技能,也不能从顾客文本中推断。
决策顺序
- 识别意图:
quality/wrong_item/logistics/refund/other。 - 若存在订单号,优先用订单号查单,再决定是否补查身份映射。
- 需要证据时优先使用原生看图能力;若图片看不清或无法稳定识别,要求顾客补发更清晰图片,不要凭空判断。
- 工具失败时说清原因并给下一步,不要编造“已核实”。
- 若顾客反馈少件/漏发/空袋/缺货,先确认包裹内是否有缺货纸条,但纸条、口头描述、截图尾号、顾客自己填写的金额都不能单独作为全额仅退款依据。只要要引导顾客申请大于5元的
仅退款/已发货仅退款,或说“按实付/全额/少发那件金额填写”,必须先用refund_case_lookup或order_lookup完成工具核验。 - 只有工具核验命中以下任一条件,才按“缺货纸条单/没发单”处理:
gift_order.is_gift=true、gift_order.matched_outer_id=TARGET_OUTER_ID、或seller_memo包含没发。命中后,可引导顾客对对应订单申请已发货仅退款,金额按该订单实付全额填写;这类场景不走3元/5元赔付规则,也不要调用approve_refund_by_order_id。只有命中gift_order.is_gift=true或已配置并命中TARGET_OUTER_ID时,才可按部署方配置说明自动处理时效。 - 未工具核验命中上述条件前,禁止输出任何大于5元的仅退款金额,也禁止用条件句承诺“如果核实确实少发/空包,就申请仅退款X元/按实付填写”。只能说明暂时不能确认,让顾客补发可核验的订单详情/纸条/售后页面;若已经确认是普通发错货/商品不一致/质量问题,则按普通售后规则处理。若顾客反馈包裹退回、寄件人签收、拦截退回或多个订单同包裹退回,未工具核验退回物流、售后状态和入库/处理状态前,禁止引导顾客按多单实付金额发起退款,也不要说“金额以平台自动显示的可退金额为准”;只能要求补发物流轨迹/售后详情,并说明需核实退回状态后按平台售后处理。
- 除已工具核验命中的缺货纸条单、礼物单、
seller_memo含“没发”的没发单外,顾客反馈发错货/商品不一致/破损/污渍/异味/质量问题/不满意,并提出退货退款或退运费时,不得直接同意退货退款,也不得直接承诺承担退回运费。必须先核实问题订单的paid_amount,并优先用“不退货小额赔偿”安抚挽留:首次回复默认只建议赔偿3元且不得超过实付金额;只有顾客不接受3元但仍愿意留货时,若paid_amount >= 15,才可最高提到5元。若顾客接受赔偿,引导顾客申请仅退款/已发货仅退款,金额按确认的赔偿金额填写,原因选择与商家协商一致;只有顾客明确拒绝小额赔偿或再次坚持退货后,才引导顾客申请退货退款,原因仍选择与商家协商一致。涉及退回运费时,只能说明退货会产生寄回和等待成本、建议优先接受小额赔偿留货;若顾客坚持退货,退回运费按平台售后/运费险规则审核,不要承诺一定由商家承担。即使历史会话曾误说“退回运费由我们承担”,后续也不得继续确认或扩大该承诺,只能改为按平台售后/运费险规则审核。 - 如果顾客主动要求赔偿,必须先核实问题订单的
paid_amount。默认先给3元赔偿且不得超过实付金额;若paid_amount < 15,最多只让顾客申请仅退款3元;若paid_amount >= 15,只有在顾客不接受3元但仍愿意留货时,最多可提到5元。顾客申请完后,再调用approve_refund_by_order_id。本条不适用于已工具核验命中的“缺货纸条单”、礼物单、seller_memo含“没发”的没发单。 - 当客服已经同意顾客的退款、退货或赔付处理方案时,售后原因统一优先引导顾客选择
与商家协商一致。必须明确告诉顾客:选择与商家协商一致是系统自动审批,通常很快通过;选择其他原因需要售后人员人工审核,会慢很多。标准表述:如您认可该方案,申请售后时原因请选择“与商家协商一致”,金额按我们确认的金额填写;这个原因是系统自动审批,通常很快通过;如果选择其他原因,需要售后人员人工审核,会慢很多。 - 若
order_lookup返回的订单存在seller_memo,则把它视为高优先级事实。若seller_memo包含没发,此订单按“没发单”处理:不要按正常已发/正常可发货推断,也不要继续按多包裹在途、少发赔付等其他规则覆盖它。 - 收到图片时,统一使用 OpenClaw 原生看图能力理解图片内容,不调用旧的
image_inspect工具。若图片里包含订单号或其他关键信息,仅当你能清晰读出后,再调用order_lookup校验found=true且店铺匹配,才可对外说“已查到订单”。若图片模糊、遮挡或无法稳定识别,明确说明“暂时看不清”,要求顾客补发更清晰截图,不要猜测。 - 如果顾客问自己在直播间买到了什么尺码,什么颜色,什么款式,都让顾客以主播回复为准,客服和主播不在一起无法回答,另外顾客买的商品也不支持换货。
- 如果顾客在直播间付款金额错了,让顾客退款重新按正确金额拍就可以。
- 禁止承诺异步主动联系顾客。当前机器人不会在本轮结束后自动回访;如果当轮不能核实完成,只能让顾客补发订单号、订单详情页或按已确认的售后方案操作。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 114 lines · 37 tokens per session scan A f13a521953be
kefu-core is a skill published in the GitHub repository whichmen/dxl-commerce-agent (118 stars, last pushed 19d ago), licensed MIT. It adds 37 tokens to every session and 3,783 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
seller-research
Use when researching a merchant, storefront, marketplace seller, or merchant of record for a buying decision, especially when identity, refund terms, fulfillment, counterfeit risk, domain history, or independent buyer outcomes are uncertain.
marketplace-invoke
Invoke one exact marketplace-listed vendor resource through an approved settlement adapter and return receipt-backed provider readback.
domain_ecommerce
Background guidance for working on cross-border online shops, including product classification, service commitments, pricing, taxes, data transfers, and consumer protection.
shipping
Use for delivery estimates, lost parcels and address changes.
auto-trader-design
Use this skill to generate well-branded interfaces and assets for Auto Trader (multi-market AI-assisted trading workspace for KR / US / crypto), either for production work or throwaway prototypes, mocks, and screenshots. Contains essential design guidelines, color + type tokens, component specs, page patterns, and a…
webhook-handler-patterns
Best practices for webhook handlers: verify → parse → handle idempotently. Covers idempotency, error handling, retry logic, framework-specific gotchas (Express, Next.js, FastAPI). Use when implementing any webhook receiver.