‹ 首页

escalation-design

@owl-listener · 收录于 5 天前 · 上游提交 1 个月前

When and how AI should escalate to humans, refuse, or ask for clarification.

场景ai行为设计人机协作

适合你，如果你在构建AI系统时需要定义何时求助人类。

/ 下载安装
https://github.com/owl-listener/ai-design-skills/tree/main/claude-plugin/ai-alignment-reasoning/skills/escalation-design ↗
/ 通过 npx 安装 校验哈希
npx oh-my-skill add owl-listener/ai-design-skills/escalation-design
/ 通过 bash 安装
curl -fsSL https://oh-my-skill.com/install.sh | bash -s -- owl-listener/ai-design-skills/escalation-design
/ 已经装过?验证本机副本,不用重装
npx oh-my-skill verify owl-listener/ai-design-skills/escalation-design
安装目标可用 --agent / --scope 或 --to 明确指定;省略时只会在唯一已存在的 agent 目录上自动选择,零命中或多命中会停止并提示。content_hash 缺失或不一致均拒装。
144GitHub stars
~553上下文体积 · 单文件
索引托管

怎么用

商店整理自技能原文 · 版本 f41b650 · 表述以原文为准
它做什么

本技能指导AI在遇到不确定、高风险、用户情绪波动等情况时，主动将问题升级给人工客服或建议用户自行决策，而不是死循环或擅自处理。

什么时候触发

当AI对回答信心低、任务涉及重大后果（如财务、医疗）、用户表现出情绪困扰、请求超出范围或违反策略，以及无法澄清歧义时触发。

装好后可以这样说
AI识别到信心不足时，会主动请求人工介入。
高风险场景下，AI拒绝并指引正确渠道。
检测到用户情绪 distress 时，提供升级选项。
技能原文 SKILL.md作者撰写 · MIT · f41b650

Escalation Design

Escalation is what happens when the AI reaches the boundary of what it should handle alone. Designing escalation well means the user gets help instead of a dead end — and the AI knows its limits.

Escalation Triggers

The AI should escalate when:

  • Confidence is low: The AI isn't sure its output is correct or helpful
  • Stakes are high: The decision has significant consequences (financial, medical, legal, safety)
  • Emotional distress: The user shows signs of crisis, distress, or vulnerability
  • Ambiguity is unresolvable: The AI can't determine intent even after clarification
  • Scope boundary: The request is outside what the AI is designed to handle
  • Policy boundary: The request approaches or crosses a guardrail
  • Conflict: The user disagrees with the AI and the disagreement can't be resolved
Escalation Types
  • To human support: Transfer to a human agent with full context
  • To the user themselves: "This decision is yours to make" — handing back agency
  • To a specialist: Routing to domain-specific help (medical, legal, technical)
  • To a supervisor/admin: Flagging for organisational review
  • Self-escalation: The AI flags its own output for review before delivering it
Designing the Escalation Experience

The user's experience of escalation matters:

  • Context transfer: When escalating to a human, pass the full conversation. Don't make the user repeat themselves.
  • Warm handoff: "I'm connecting you with someone who can help with this" — not a cold redirect.
  • Expectation setting: Tell the user what will happen next and how long it might take.
  • Graceful degradation: If no human is available, offer alternatives — not a dead end.
  • Dignity: Never make the user feel stupid for needing escalation.
Escalation Anti-Patterns
  • The infinite loop: AI keeps trying instead of escalating, frustrating the user
  • Premature escalation: AI escalates when it could easily handle the request, annoying the user
  • Context loss: User has to start over after escalation
  • Blame shifting: AI implies the user caused the problem
  • Hidden escalation: Escalation happens without the user knowing
Design Artefacts
  • Escalation trigger matrix: Trigger | Threshold | Escalation Type | User Experience
  • Escalation flow diagrams per feature
  • Context handoff specifications
  • Fallback path designs for when escalation isn't available
  • Escalation quality metrics
按 MIT 许可原样转载,未经改动 · 在 GitHub 查看 →

评论

登录即可评论;带「已验证安装」的,是发布者名下有本店的安装或持有记录。
…