← 返回 Skills 市场

Complaint Handler

Name: Complaint Handler
Author: fangwei-frank

作者 fangwei-frank · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

102

总下载

当前安装

版本数

在 OpenClaw 中安装

/install complaint-handler

功能描述

Retail complaint and after-sales handler for digital employees. Classifies complaints, generates empathetic responses, routes escalations, and manages return...

使用说明 (SKILL.md)

Complaint Handler

Overview

This skill manages negative customer interactions: complaints, quality issues, return/exchange requests, and escalations. Its job is to de-escalate, resolve what it can, and route what it can't — always within the configured permission matrix.

Depends on: policy_entries in knowledge base + permissions_config from Step 09.

Complaint Classification

Classify every incoming complaint before responding:

Class	Trigger	Default Level	Example
`quality_issue`	质量/坏/破损/开线/褪色/异味	L1	"衣服洗了之后褪色了"
`size_fit_issue`	尺寸/不合适/太大/太小	L0	"买的M码穿着偏大"
`wrong_item`	发错/和描述不符/不是我要的	L1	"收到的颜色不对"
`refund_request`	退款/退钱/要退	L1/L2	"我要退款"
`exchange_request`	换货/换一个/换个码	L1	"能不能给我换个L码"
`service_complaint`	态度/服务/等太久	L0	"你们员工态度很差"
`escalation_threat`	律师/媒体/消协/曝光/投诉到	L3	"我要找消费者协会"
`abuse`	辱骂/人身攻击	L3	[profanity detected]

Reference: classification-guide.md

Response Protocol

Step 1: Acknowledge (always first)

Never jump to solutions without acknowledging the customer's frustration. Template: "非常抱歉给您带来不便，我完全理解您现在的感受。" Adjust warmth based on severity: mild issue → warm; strong emotion → deeply empathetic.

Step 2: Clarify (if needed)

Ask one targeted question to understand the situation:

"请问是什么时候购买的呢？"
"方便描述一下具体是什么问题吗？" Never ask multiple questions at once.

Step 3: Apply Policy

Look up the relevant policy from knowledge base. Apply exactly.

State what the customer is entitled to (specific, no vague promises)
State the conditions they need to meet
State the next step clearly

Step 4: Execute or Escalate

L0: Handle fully, confirm resolution
L1: Propose action, wait for staff confirmation tap
L2: Create ticket, notify manager, give customer ETA
L3: Immediately hand off to human, stay on standby

Reference: response-templates.md

Escalation Triggers (Auto L3)

Always escalate to L3 immediately on detection of:

Legal keywords: 律师, 法院, 起诉, 法律途径
Media keywords: 媒体, 曝光, 记者, 微博, 抖音发
Authority keywords: 消协, 12315, 工商, 监管
Repeated contact: same issue raised 3+ times
Explicit threat: 骗子, 假货, 虚假宣传
Abuse: profanity or personal attacks

On L3 trigger:

Stop trying to resolve
Acknowledge and transfer: "您的情况非常重要，我马上为您转接专属客服，请稍候。"
Send escalation packet to L3 contact (see permissions_config)
Do NOT argue, defend, or explain further

What This Skill Will Never Do

Promise a specific refund amount without human approval
Approve a refund exceeding refund_auto_approve_limit (default: 0)
Commit to a pickup/exchange date without system confirmation
Blame staff members by name
Deny a clearly valid claim to avoid a refund
Claim the customer is wrong about a factual quality issue

安全使用建议

This skill looks like a plausible customer-support assistant, but it relies on external policy/configuration and escalation channels that are not declared in the registry. Before installing or enabling it: 1) Ensure a documented permissions_config and policy_entries KB exist and are reachable, and add those config paths or required env vars to the skill metadata; 2) Confirm exactly where 'escalation packets' will be sent, what authentication is used, and which fields (PII) are included; 3) Limit the agent's rights to only the necessary systems (ticketing, manager contacts) and log all escalations for audit; 4) Test in a sandbox with redacted data to verify the handoff behavior; 5) If you cannot control or verify the escalation endpoint and auth, treat the skill as risky because it could transmit customer data to an unknown destination.

功能分析

Type: OpenClaw Skill Name: complaint-handler Version: 1.0.0 The 'complaint-handler' skill is a standard customer service automation tool designed to classify and respond to retail complaints. It includes well-defined logic for escalation (L0-L3) based on keywords related to legal threats or media exposure, and provides structured response templates. The skill contains explicit safety constraints, such as prohibiting unauthorized refunds and preventing the agent from blaming specific staff members. No indicators of data exfiltration, malicious execution, or harmful prompt injection were found across SKILL.md or its reference files.

能力评估

⚠ Purpose & Capability

The SKILL.md explicitly depends on 'policy_entries' in a knowledge base and a 'permissions_config' (used to look up escalation contacts, approval thresholds, and packet format), but the registry metadata declares no required config paths, environment variables, or primary credential. That mismatch means the skill expects access to external policy/configuration data without declaring how or where those secrets/endpoints live.

⚠ Instruction Scope

Instructions require collecting PII (order screenshots, addresses, contact info) and sending 'escalation packets' to L2/L3 contacts, but they do not specify destinations, authentication, transport, or what fields to include beyond generic phrasing. The two-pass classification and keyword-based escalation are coherent with the purpose, but the handoff/transmit steps grant wide discretion to the agent about where and how to send potentially sensitive customer data.

✓ Install Mechanism

This is instruction-only with no install spec and no code files. That minimizes supply-chain risk — nothing is downloaded or installed by the skill itself.

⚠ Credentials

No environment variables or config paths are declared even though the instructions require permissions_config (escalation contacts, thresholds) and access to a policy knowledge base. The absence of declared credentials/configs is disproportionate to the stated need and leaves ambiguity about what system privileges or secrets the runtime agent must use to comply with the skill.

✓ Persistence & Privilege

The skill does not request always:true or other elevated persistent privileges and does not instruct modifying other skills or global agent configuration. Autonomous invocation is allowed by default but is not, on its own, flagged.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install complaint-handler
安装完成后，直接呼叫该 Skill 的名称或使用 /complaint-handler 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

first release

元数据

Slug complaint-handler

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题