← 返回 Skills 市场
Chaos Cognitive Resilience
作者
KingOfZhao
· GitHub ↗
· v1.0.0
· MIT-0
124
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install chaos-cognitive-resilience
功能描述
Design and test cognitive systems to gracefully recover from errors and overloads using chaos engineering principles and resilience metrics.
使用说明 (SKILL.md)
Chaos Cognitive Resilience
混沌工程 × 认知科学:在不确定性中构建弹性认知系统。
何时使用
- 设计能在信息噪声、对抗性输入、部分故障下仍正常工作的认知系统
- 评估AI系统在认知过载场景下的鲁棒性
- 构建人类-AI协作中的认知韧性
核心认知
1. 认知系统的混沌面
传统混沌工程关注基础设施(网络延迟、节点故障)。认知系统的混沌面完全不同:
- 信息混沌:输入数据包含错误、矛盾、缺失
- 注意力混沌:同时处理多个竞争性任务导致认知过载
- 记忆混沌:长期记忆退化、短期记忆溢出、记忆冲突
- 推理混沌:部分前提错误导致推理链崩溃
认知系统不需要"不犯错",而是需要"犯错后优雅恢复"。这与传统混沌工程的"blast radius控制"概念完美对应。
2. 认知混沌猴子(Cognitive Chaos Monkey)
类似Netflix的Chaos Monkey,但注入的是认知干扰:
- 语义翻转:随机将输入中的关键词替换为反义词
- 上下文截断:模拟注意力窗口不足,截断关键上下文
- 记忆注入:插入虚假"历史记忆",测试系统辨别能力
- 推理分支炸弹:注入需要指数级推理时间的输入
关键指标不是"是否出错",而是"错误后的恢复路径质量和时间"。
3. 认知弹性指数(Cognitive Resilience Index, CRI)
CRI = Σ(恢复速度_i × 恢复质量_i × 权重_i) / N
其中:
- 恢复速度:从错误状态恢复到正常状态的时间
- 恢复质量:恢复后的认知输出准确度
- 权重:该故障模式的重要程度
- N:测试的故障模式数量
实践框架
Phase 1: 认知故障模式枚举
├── 输入层故障(噪声、缺失、矛盾、对抗)
├── 处理层故障(注意力不足、推理链断裂、偏见放大)
├── 输出层故障(幻觉、过度自信、决策瘫痪)
└── 系统层故障(记忆溢出、上下文丢失、模型切换异常)
Phase 2: 混沌注入实验
├── 定义稳态基线(正常状态下的输出质量)
├── 注入单一故障 → 测量偏差
├── 注入组合故障 → 测量交互效应
└── 渐进加大故障强度 → 找到系统崩溃临界点
Phase 3: 弹性增强
├── 输入校验层(异常检测 → 置信度标注 → 降级处理)
├── 推理冗余层(多路径推理 → 交叉验证 → 共识机制)
├── 输出保护层(事实核查 → 一致性检查 → 不确定性表达)
└元认知监控层(自我评估 → 异常检测 → 主动降级)
Phase 4: 持续验证
├── 自动化混沌测试管线
├── CRI趋势追踪
└── 新故障模式发现与归类
碰撞来源
chaos-engineering-playbook×cognitive-load-testing×cognitive-bias-in-engineeringself-healing-database(恢复模式)×neuroscience-for-devs(认知模型)
安全使用建议
This is a conceptual playbook (no code, no installs, no credentials) and is coherent with its stated purpose. Risks are operational rather than immediate: if you implement its experiments, run chaos injections only in isolated/test environments, avoid exposing real user data or PII, limit blast radius of tests, and add monitoring/kill-switches. Also note the skill's source/homepage are unknown — treat it as third‑party guidance and review any concrete test harness or scripts you build from it before running in production.
功能分析
Type: OpenClaw Skill
Name: chaos-cognitive-resilience
Version: 1.0.0
The skill bundle contains purely conceptual and theoretical documentation regarding 'Chaos Cognitive Resilience,' applying chaos engineering principles to AI systems. There is no executable code, shell commands, or instructions that would direct an agent to perform unauthorized actions or access sensitive data. The content in SKILL.md is educational and lacks any indicators of malicious intent or technical vulnerabilities.
能力评估
Purpose & Capability
The name and description (cognitive resilience via chaos engineering) match the SKILL.md content. No unexpected binaries, env vars, or config paths are requested. The lack of install or credentials is proportionate for a guidance/playbook.
Instruction Scope
SKILL.md contains conceptual frameworks, metrics (CRI), and procedural phases for testing and hardening cognitive systems. It does not instruct the agent to read system files, access environment variables, call external endpoints, or execute commands. There is no scope creep in the written instructions.
Install Mechanism
There is no install spec and no code files. Because this is instruction-only, nothing is written to disk or downloaded by the skill itself.
Credentials
The skill requests no environment variables, credentials, or config paths. That is appropriate for a conceptual playbook; there are no disproportionate or unexplained secrets requested.
Persistence & Privilege
always is false and the skill does not request permanent presence or privileged modifications. Autonomous invocation is allowed by platform default but this skill's content does not enable privileged actions.
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 chaos-cognitive-resilience - 安装完成后,直接呼叫该 Skill 的名称或使用
/chaos-cognitive-resilience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
Initial release introducing the core framework for building cognitive resilience through chaos engineering concepts.
- Defines cognitive failure modes such as information, attention, memory, and reasoning chaos.
- Introduces "Cognitive Chaos Monkey" for injecting realistic cognitive faults.
- Proposes the Cognitive Resilience Index (CRI) to measure recovery speed and quality after errors.
- Outlines a four-phase practical framework: fault enumeration, chaos injection, resilience enhancement, and continuous validation.
- Integrates ideas from chaos engineering, cognitive load testing, bias analysis, and neuroscientific models.
元数据
常见问题
Chaos Cognitive Resilience 是什么?
Design and test cognitive systems to gracefully recover from errors and overloads using chaos engineering principles and resilience metrics. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 124 次。
如何安装 Chaos Cognitive Resilience?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install chaos-cognitive-resilience」即可一键安装,无需额外配置。
Chaos Cognitive Resilience 是免费的吗?
是的,Chaos Cognitive Resilience 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Chaos Cognitive Resilience 支持哪些平台?
Chaos Cognitive Resilience 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 Chaos Cognitive Resilience?
由 KingOfZhao(@kingofzhao)开发并维护,当前版本 v1.0.0。
推荐 Skills