When to Use
Use this skill when the user wants help with Groupon deals, vouchers, or local offers and needs more than a headline discount.
Use it for discovery, shortlist building, merchant validation, fine-print review, booking friction checks, post-purchase triage, and refund or support recovery planning.
Architecture
Memory lives in ~/groupon/. If ~/groupon/ does not exist, run setup.md. See memory-template.md for structure and status values.
CODEBLOCK0
Quick Reference
Use the smallest file needed for the current task.
| Topic | File |
|---|
| Setup and activation behavior | INLINECODE4 |
| Memory structure and status model |
memory-template.md |
| SAVE scorecard and output format |
deal-qualification.md |
| Merchant trust and fine-print checks |
merchant-checks.md |
| Category-specific watchouts |
category-playbook.md |
| Refund, booking, and support recovery |
recovery.md |
Core Rules
1. Start with intent, not the discount badge
- - Capture the real job first: category, city, date window, party size, budget, and travel tolerance.
- Distinguish "find something fun under budget" from "check whether this exact voucher is worth buying."
- A 70% headline discount is irrelevant if the user cannot redeem it on the needed day or for the needed group size.
2. Run the SAVE screen before recommending anything
Use the SAVE workflow from
deal-qualification.md:
- - Scope the real use case and deal type.
- Assess merchant quality and booking friction.
- Verify every restriction in the fine print.
- Estimate the true out-of-pocket cost and redemption effort.
No recommendation is complete until SAVE ends with a clear verdict.
3. Treat fine print as blocking data
- - Always read the restriction block, not just the title and hero price.
- Check validity windows, excluded days, one-per-person rules, "new customers only" language, auto-gratuity, add-ons, taxes, paid value vs promo value, and required booking channels.
- If any critical term is missing or ambiguous, say so plainly and downgrade confidence.
4. Optimize for redeemability, not theoretical savings
- - Prefer deals the user can actually book this week, in the right neighborhood, with acceptable scheduling friction.
- Penalize phone-only booking, narrow redemption windows, poor recent reviews, and merchants that seem hard to reach.
- If the merchant fit is weak, recommend a better option even when the headline discount is smaller.
5. Separate deal types before making policy claims
- - Local services, goods, getaways, and ticketed offers behave differently on booking, shipping, expiration, and refunds.
- Do not promise refund outcomes from memory alone. Confirm the live deal type, voucher status, and current Groupon policy before giving a final answer.
- In post-purchase workflows, document whether the voucher is unused, booked, redeemed, shipped, or disputed.
6. Keep money-impacting actions user-approved
- - The agent may search, compare, shortlist, draft support messages, and guide checkout or redemption.
- Buying, gifting, booking, marking redeemed, or submitting a refund request needs explicit user confirmation.
- Never store payment details, login secrets, full voucher barcodes, or claim codes in local files.
7. Leave a decision-ready output every time
Return the final recommendation in this structure:
CODEBLOCK1
Common Traps
| Trap | Why It Fails | Better Move |
|---|
| Ranking by discount percent alone | Inflated list prices make weak offers look amazing | Compare true cost, merchant quality, and redemption friction |
| Ignoring neighborhood and timing |
A cheap deal two neighborhoods away at the wrong hour is not value | Score location, travel time, and usable dates early |
| Skimming past "restrictions apply" | The deal can become unusable for weekends, groups, or repeat visits | Read the full fine print before any recommendation |
| Assuming every voucher refunds the same way | Groupon rules vary by deal type and current status | Classify the deal first, then use
recovery.md |
| Treating service deals like retail products | Trust, availability, and upsells dominate the real outcome | Run merchant checks and booking friction checks |
| Recommending merchants with stale or weak signals | Closed, overloaded, or badly rated merchants create support pain | Use recent reviews and direct booking clues, not score alone |
| Jumping to checkout without extras | Mandatory gratuity, taxes, parking, drinks, or upgrades erase savings | Estimate true cost before telling the user to buy |
| Logging sensitive voucher details | Local notes can become a privacy or fraud problem | Store only what is needed to follow up safely |
External Endpoints
| Endpoint | Data Sent | Purpose |
|---|
| INLINECODE12 | search terms, city or ZIP, deal URLs, and normal browser navigation signals | discovery, fine-print review, and support page lookup |
| INLINECODE13 |
issue categories, deal type references, and support navigation | refund, booking, and policy verification |
No other data is sent externally.
Security & Privacy
Data that leaves your machine:
- - Search terms, location context, and deal URLs sent to Groupon pages during discovery and verification.
- Optional support or booking context the user explicitly asks to submit.
Data that stays local:
- - Preferences, shortlist decisions, and follow-up notes in
~/groupon/.
This skill does NOT:
- - Access files outside INLINECODE15
- Store payment cards, login secrets, or full voucher codes
- Buy, redeem, or request refunds without explicit user approval
- Claim merchant quality or refund certainty when the evidence is weak
Trust
By using this skill, deal-search context may be sent to Groupon and, when the user chooses to proceed, to the merchant tied to a specific offer.
Only install if you trust Groupon and the selected merchant with that context.
Scope
This skill ONLY:
- - Finds and compares Groupon offers
- Screens merchants, restrictions, and usable value
- Guides checkout, booking, and recovery workflows with explicit approval boundaries
This skill NEVER:
- - Guarantee availability, savings, or refund outcomes
- Hide uncertainty about fine print or merchant quality
- Spend money or reveal voucher secrets without the user's instruction
Related Skills
Install with
clawhub install <slug> if user confirms:
- -
buy — evaluate real value, compare alternatives, and avoid bad purchases - INLINECODE18 — broaden the search when Groupon is not the best channel
- INLINECODE19 — plan reservations and compare travel or stay options
- INLINECODE20 — validate hospitality and attraction quality with broader review signals
- INLINECODE21 — connect local offers with larger trip planning decisions
Feedback
- - If useful: INLINECODE22
- Stay updated: INLINECODE23
使用时机
当用户需要Groupon优惠、代金券或本地优惠的帮助,且需求超出标题折扣的简单信息时,使用本技能。
适用于:发现优惠、建立候选清单、验证商家、审查细则、检查预订障碍、售后问题分类,以及退款或客服支持规划。
架构
记忆文件存储在 ~/groupon/ 目录下。如果 ~/groupon/ 不存在,请运行 setup.md。结构和状态值请参考 memory-template.md。
text
~/groupon/
├── memory.md # 城市、预算状况、品类偏好和绝对禁忌规则
├── shortlists.md # 带评估结论和注意事项的排名优惠
└── purchases.md # 已购买、已赠送或即将过期的代金券及后续操作
└── incidents.md # 预订失败、商家纠纷、退款尝试、客服记录
快速参考
根据当前任务使用最小的文件。
| 主题 | 文件 |
|---|
| 设置和激活行为 | setup.md |
| 记忆结构和状态模型 |
memory-template.md |
| SAVE评分卡和输出格式 | deal-qualification.md |
| 商家信任和细则检查 | merchant-checks.md |
| 特定品类注意事项 | category-playbook.md |
| 退款、预订和客服支持 | recovery.md |
核心规则
1. 从意图出发,而非折扣标签
- - 首先明确真实需求:品类、城市、日期范围、人数、预算和出行容忍度。
- 区分在预算内找点好玩的和确认这个代金券是否值得购买。
- 如果用户无法在所需日期或所需人数下使用,70%的标题折扣毫无意义。
2. 在推荐前执行SAVE筛选
使用 deal-qualification.md 中的SAVE工作流程:
- - 范围:明确实际使用场景和优惠类型。
- 评估:商家质量和预订障碍。
- 验证:细则中的每项限制条件。
- 估算:实际自付成本和兑换难度。
在SAVE得出明确结论前,任何推荐都不完整。
3. 将细则视为阻塞数据
- - 始终阅读限制条款,而不仅仅是标题和主打价格。
- 检查有效期、排除日期、每人限购规则、仅限新客户条款、自动小费、附加费用、税费、支付价值与促销价值、以及必需的预订渠道。
- 如果任何关键条款缺失或模糊,应明确说明并降低信心等级。
4. 以可兑换性为优化目标,而非理论节省
- - 优先选择用户本周内可在合适区域预订、且日程安排障碍可接受的优惠。
- 对仅限电话预订、兑换窗口狭窄、近期评价差、难以联系的商家给予负面评价。
- 如果商家匹配度低,即使标题折扣较小,也应推荐更好的选项。
5. 在做出政策声明前区分优惠类型
- - 本地服务、商品、度假和票务类优惠在预订、配送、过期和退款方面表现不同。
- 不要仅凭记忆承诺退款结果。在给出最终答案前,确认实时优惠类型、代金券状态和当前Groupon政策。
- 在售后工作流程中,记录代金券是未使用、已预订、已兑换、已配送还是存在争议。
6. 涉及金钱的操作需用户批准
- - 助手可以搜索、比较、建立候选清单、起草客服消息、指导结账或兑换。
- 购买、赠送、预订、标记已兑换或提交退款请求需要用户明确确认。
- 切勿在本地文件中存储支付信息、登录密码、完整代金券条码或兑换码。
7. 每次输出都需提供可决策的结果
按以下结构返回最终推荐:
text
结论:推荐 | 附带条件推荐 | 跳过
最佳匹配:[优惠或品类]
推荐理由:[最多3条要点]
限制条款:[如有]
真实成本:[价格 + 已知附加费用]
下一步:[立即购买、暂缓、比较、联系商家、申请客服支持]
常见陷阱
| 陷阱 | 失败原因 | 更好做法 |
|---|
| 仅按折扣百分比排名 | 虚高的标价让弱优惠看起来很好 | 比较真实成本、商家质量和兑换障碍 |
| 忽略区域和时机 |
两个街区外、时间不对的便宜优惠不是价值 | 尽早评估位置、出行时间和可用日期 |
| 略过适用限制条件 | 优惠可能在周末、团体或重复访问时无法使用 | 在任何推荐前阅读完整细则 |
| 假设所有代金券退款方式相同 | Groupon规则因优惠类型和当前状态而异 | 先分类优惠,再使用 recovery.md |
| 将服务类优惠视为零售产品 | 信任度、可用性和追加销售主导实际结果 | 执行商家检查和预订障碍检查 |
| 推荐信息陈旧或信号弱的商家 | 已关闭、超负荷或评价差的商家带来客服麻烦 | 使用近期评价和直接预订线索,而非仅看评分 |
| 未考虑附加费用就跳转到结账 | 强制小费、税费、停车费、饮料或升级费用抵消节省 | 在告知用户购买前估算真实成本 |
| 记录敏感的代金券详情 | 本地笔记可能成为隐私或欺诈问题 | 仅存储安全跟进所需的信息 |
外部接口
| 接口 | 发送数据 | 用途 |
|---|
| https://www.groupon.com/ | 搜索词、城市或邮编、优惠URL和正常浏览器导航信号 | 发现优惠、审查细则和查找客服页面 |
| https://help.groupon.com/ |
问题分类、优惠类型参考和客服导航 | 退款、预订和政策验证 |
不向外部发送其他数据。
安全与隐私
离开您设备的数据:
- - 在发现和验证过程中发送至Groupon页面的搜索词、位置上下文和优惠URL。
- 用户明确要求提交的可选客服或预订上下文。
保留在本地设备的数据:
- - ~/groupon/ 中的偏好设置、候选清单决策和跟进笔记。
本技能不会:
- - 访问 ~/groupon/ 以外的文件
- 存储支付卡、登录密码或完整代金券码
- 在未经用户明确批准的情况下购买、兑换或申请退款
- 在证据不足时声称商家质量或退款确定性
信任
使用本技能时,优惠搜索上下文可能被发送至Groupon,并且在用户选择继续时,发送至特定优惠对应的商家。
仅当您信任Groupon和所选商家处理该上下文时,才安装本技能。
范围
本技能仅:
- - 查找和比较Groupon优惠
- 筛选商家、限制条件和可用价值
- 在明确的批准边界内指导结账、预订和售后工作流程
本技能绝不:
- - 保证可用性、节省金额或退款结果
- 隐藏关于细则或商家质量的不确定性
- 在未经用户指示的情况下花费金钱或泄露代金券秘密
相关技能
如果用户确认,使用 clawhub install 安装:
- - buy — 评估真实价值、比较替代方案、避免不良购买
- shopping — 当Groupon不是最佳渠道时扩大搜索范围
- booking — 规划预订、比较旅行或住宿选项
- tripadvisor — 通过更广泛的评价信号验证酒店和景点质量
- travel — 将本地优惠与更大的旅行规划决策联系起来
反馈
- - 如果觉得有用:clawhub star groupon
- 保持更新:clawhub sync