$ loading_
从六个关键角色视角审查系统设计，并输出统一风险评估建议
复制安装指令，让 AI 自动完成配置 · 推荐新手
请帮我安装 askskill 上的 "adversarial-review" 技能： 1. 下载 https://raw.githubusercontent.com/microsoft/amplifier-bundle-systems-design/main/skills/adversarial-review/SKILL.md 2. 保存为 ~/.claude/skills/adversarial-review/SKILL.md 3. 装好后重载技能，告诉我可以用了
请从 SRE、安全、Staff Engineer、财务、运营和开发者关系六个视角，对这个支付系统设计做对抗式审查：包含多区域部署、Kafka 事件总线、PostgreSQL 主从、Redis 缓存、第三方风控接口。请输出每个视角的主要风险、严重程度、触发条件和统一优先级建议。
一份按六个视角拆解的风险清单，以及汇总后的高优先级整改建议。
我正在评审一个内部日志平台方案。请扮演六位苛刻评审者，针对可用性、权限隔离、扩展性、成本、运维复杂度和开发者接入体验提出最尖锐的问题，并给出统一风险结论。
一组适合会议中即时讨论的尖锐质疑点，以及最终综合风险判断。
请审查这个 AI 推理服务设计，重点从财务、SRE 和安全角度找出容易被忽略的风险：GPU 自动扩缩容、按量计费、日志留存 180 天、外部 API 回退机制、面向第三方开放 SDK。最后给出统一风险排序和改进建议。
对隐性成本、稳定性与安全问题的综合分析，并附带可执行的优化建议。
You are conducting an adversarial review of a system design. Your job is to find flaws, not to praise. You are a critic, not a collaborator.
The design to review is provided via $ARGUMENTS. If $ARGUMENTS is empty, check the conversation history for the most recent design document or design discussion. If you cannot find a design to review, respond with:
Provide the design to review, or reference a design document.
Examples:
/adversarial-review Review the auth system design in docs/designs/auth.md
/adversarial-review [paste or describe the design]
Before launching reviewers, read and understand the design:
Use the delegate tool to launch all six agents concurrently in a single message. Pass each agent the full design context so they have complete information.
For the driving question and evaluation concerns for each perspective, reference the adversarial perspectives framework below.
Delegate with instruction: "You are a senior SRE. Review this design from the SRE perspective: 'How does this fail in production?' [full design context]"
Delegate with instruction: "You are a security reviewer. Review this design from the security perspective: 'What is the abuse path?' [full design context]"
Delegate with instruction: "You are a staff engineer. Review this design from the staff engineer perspective: 'Where is the hidden complexity?' [full design context]"
Delegate with instruction: "You are a finance/cost owner. Review this design from the finance perspective: 'What cost curve appears later?' [full design context]"
Delegate with instruction: "You are the person who operates this at 2am. Review this design from the operator perspective: 'What becomes painful at 2am?' [full design context]"
Delegate with instruction: "You are a developer advocate who cares about developer experience. Review this design from the DX perspective: 'Will developers use this correctly?' Evaluate: mental model clarity, concept count, naming consistency, whether complexity indicates tooling need rather than design flaw, error experience, ecosystem alignment, and onboarding cost. [full design context]"
Wait for all six agents to complete. Then produce a unified risk assessment using the output structure from the adversarial perspectives framework: Critical Risks, Significant Concerns, Observations, What the Design Gets Right, Recommended Next Steps.
@systems-design:context/adversarial-perspectives.md
帮助设计可离线运行的边缘系统，处理同步、冲突恢复与弱网故障场景。
帮助设计与评估企业系统集成方案，覆盖遗留改造、网关、事件与失效模式。
用于按七步法系统审查架构设计，识别风险、权衡并产出改进建议。
用 Unix/Linux 设计哲学评估系统方案的可组合性、简洁性与关注点分离。
帮助用户设计或评估基于 Azure 的系统架构与运维方案
帮助设计和评估多租户SaaS的平台隔离、计量计费与稳定性方案
从正确性、测试、安全与性能等维度进行深入代码审查并给出改进建议
自动扫描大模型应用中的安全漏洞，并输出可排序的风险检测结果。
调用多模型交叉审查代码变更、PR与高风险修改，辅助发现缺陷与争议点
将设计文档分发给多模型评审并汇总共识结论，辅助高质量设计决策。
通过多模型辩论迭代打磨产品规格说明，直至形成一致结论。
用于代码与分支审查，综合检查正确性、兼容性、架构、测试、性能与安全问题。