Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Adaptive Problem Solver

v1.0.0

Detects rigid security or permission limits and generates multiple adaptive, safe, and user-tailored alternative solutions for problem-solving.

0· 218·2 current·2 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for zs15600770520/adaptive-problem-solver.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Adaptive Problem Solver" (zs15600770520/adaptive-problem-solver) from ClawHub.
Skill page: https://clawhub.ai/zs15600770520/adaptive-problem-solver
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install adaptive-problem-solver

ClawHub CLI

Package manager switcher

npx clawhub@latest install adaptive-problem-solver
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The name/description match the implementation: code and docs focus on detecting rigid responses and generating alternative, user-facing solutions (guides, scripts, options). There are no unexpected credentials, system-level config paths, or unrelated binaries requested.
Instruction Scope
SKILL.md and the code instruct the agent to detect rigid responses, generate solution options (including host commands, container scripts, one‑click scripts, and step-by-step guides), present them to the user, and record outcomes. This stays within the skill’s purpose, but it intentionally produces commands and guidance that may ask users to run privileged or host-level operations — the skill does not auto-execute those commands, it presents them for user approval.
Install Mechanism
No install spec is present (instruction-only from registry perspective), and included files are plain JS source bundled with the skill. There are no downloads from external URLs or install-time network fetches; nothing indicates an unreviewed remote installer.
Credentials
The skill declares no environment variables, credentials, or external endpoints. It reads/writes local files (data/*.json and logs/interventions.jsonl) to store history and user profile, which is proportional to its stated learning and personalization behavior.
Persistence & Privilege
The skill metadata does not force permanent inclusion (always:false). However, the integration wrapper default config (autoTrigger: true, minConfidence threshold) means the skill is designed to auto-intervene in the assistant flow and will write logs and update a local user-profile file. This is functionally reasonable but worth noting because it changes assistant responses automatically when installed.
Assessment
This skill appears coherent with its purpose, but before installing consider: (1) review any generated 'one‑click' scripts or host commands before running them — they may require elevated privileges and could be risky if copied indiscriminately; (2) the skill stores truncated user messages and user-profile data under the skill directory and appends to logs/interventions.jsonl — check those files for sensitive content and control who can access them; (3) the integration wrapper defaults to autoTriggering interventions — if you prefer manual control, adjust autoTrigger or minConfidence in the wrapper/config before enabling; (4) monitor for any future changes that introduce network endpoints, credential requests, or code that executes commands automatically — those would materially change the risk profile. If you want, I can list the exact files/paths the skill writes to and suggest a minimal configuration to reduce auto-intervention and local logging.

Like a lobster shell, security has layers — review code before you run it.

latestvk975taw98c7cge6g9cw48whbf1838ymp
218downloads
0stars
1versions
Updated 3h ago
v1.0.0
MIT-0

Adaptive Problem Solver Skill

🎯 技能目标

解决核心问题:AI助手在安全协议限制下缺乏变通解决问题的能力

🔍 问题识别模式

常见僵化思维场景

  1. 安全协议阻塞:OpenGuardrails检测到风险 → 立即停止
  2. 权限限制:Docker容器权限不足 → 放弃操作
  3. 用户技能缺失:用户说"不会操作" → 等待用户学习
  4. 技术复杂性:多步骤复杂操作 → 要求用户手动执行
  5. 风险评估保守:中等风险 → 过度谨慎放弃

智能识别算法

class ProblemAnalyzer {
    detectRigidThinking(situation) {
        const patterns = [
            "不能操作外部系统",
            "需要用户手动处理", 
            "安全协议禁止",
            "权限不足无法执行",
            "用户需要学习",
            "过于复杂需要简化"
        ];
        
        return patterns.some(pattern => 
            situation.response.includes(pattern) && 
            !situation.response.includes("替代方案")
        );
    }
}

🛠️ 变通解决方案库

方案类型矩阵

限制类型直接方案替代方案A替代方案B简化方案
安全协议遵守禁止创建安全指南本地模拟测试分步用户指导
权限限制放弃操作权限提升请求容器外执行用户手动执行
技术复杂要求专家自动化脚本可视化工具分步教程
用户技能等待学习简化界面视频教程远程协助

具体变通策略

1. 安全协议限制变通

问题:OpenGuardrails阻止外部系统操作 变通方案

  • A1: 创建详细的用户执行指南(步骤、命令、截图)
  • A2: 开发本地测试/模拟环境
  • A3: 提供多种技术路径选择
  • A4: 分阶段渐进式实施

2. 权限限制变通

问题:Docker容器权限不足 变通方案

  • B1: 提供宿主机执行命令
  • B2: 创建容器内用户可执行脚本
  • B3: 使用环境变量传递必要信息
  • B4: 临时权限提升方案

3. 用户技能限制变通

问题:用户说"什么操作都不会" 变通方案

  • C1: 一键脚本(复制粘贴即可用)
  • C2: 可视化网页工具
  • C3: 屏幕录制指导
  • C4: 远程桌面协助指南

4. 技术复杂性变通

问题:操作过于复杂容易出错 变通方案

  • D1: 分步骤交互式脚本
  • D2: 错误自动恢复机制
  • D3: 进度保存和恢复
  • D4: 详细错误诊断

🔧 实施框架

智能决策流程

1. 问题识别
   ↓
2. 限制分析(安全/权限/技能/技术)
   ↓
3. 方案生成(直接 + 3个替代)
   ↓
4. 风险评估(低/中/高 + 缓解措施)
   ↓  
5. 方案呈现(优缺点对比)
   ↓
6. 用户选择 + 执行
   ↓
7. 反馈学习(优化方案库)

代码实现框架

class AdaptiveProblemSolver {
    constructor() {
        this.solutionLibrary = new SolutionLibrary();
        this.riskAssessor = new RiskAssessor();
        this.userProfile = new UserProfile();
    }
    
    async solve(problem, constraints) {
        // 1. 分析限制
        const limitations = this.analyzeLimitations(problem, constraints);
        
        // 2. 生成方案
        const solutions = this.generateSolutions(problem, limitations);
        
        // 3. 风险评估
        const assessedSolutions = this.assessRisks(solutions);
        
        // 4. 个性化推荐
        const recommended = this.recommendForUser(assessedSolutions);
        
        // 5. 呈现选择
        return this.presentOptions(recommended);
    }
    
    analyzeLimitations(problem, constraints) {
        return {
            security: constraints.openGuardrails ? "高风险" : "低风险",
            permissions: this.checkPermissions(),
            userSkills: this.assessUserSkills(),
            technicalComplexity: this.estimateComplexity(problem),
            timeConstraints: constraints.timeLimit
        };
    }
}

📊 学习与优化

反馈学习机制

class LearningEngine {
    constructor() {
        this.solutionHistory = [];
        this.successRates = {};
        this.userPreferences = {};
    }
    
    recordOutcome(solution, success, userFeedback, executionTime) {
        this.solutionHistory.push({
            solution,
            success,
            feedback: userFeedback,
            time: executionTime,
            timestamp: Date.now()
        });
        
        // 更新成功率
        const solutionType = solution.type;
        if (!this.successRates[solutionType]) {
            this.successRates[solutionType] = { successes: 0, attempts: 0 };
        }
        
        this.successRates[solutionType].attempts++;
        if (success) this.successRates[solutionType].successes++;
        
        // 学习用户偏好
        if (userFeedback.preference) {
            this.userPreferences[userFeedback.preference] = 
                (this.userPreferences[userFeedback.preference] || 0) + 1;
        }
    }
    
    getOptimalSolutionType(problemType) {
        // 基于历史数据推荐最优方案类型
        const candidates = Object.entries(this.successRates)
            .filter(([type, stats]) => stats.attempts >= 3)
            .map(([type, stats]) => ({
                type,
                successRate: stats.successes / stats.attempts,
                attempts: stats.attempts
            }))
            .sort((a, b) => b.successRate - a.successRate);
        
        return candidates.length > 0 ? candidates[0].type : "balanced";
    }
}

性能指标跟踪

指标目标测量方法
方案生成时间<5秒从问题识别到方案呈现
方案多样性≥3个不同类型方案数量
用户采纳率>70%用户选择执行的比例
问题解决率>85%最终成功解决问题
用户满意度>4/5反馈评分平均值

🚀 集成到工作流

与现有系统集成

  1. OpenClaw主循环集成

    // 在每次工具调用前检查
    if (adaptiveSolver.shouldIntervene(userRequest)) {
        return adaptiveSolver.solve(userRequest);
    }
    
  2. 安全协议协同

    // 与OpenGuardrails协作而非对抗
    if (openGuardrails.detected) {
        // 不是简单停止,而是触发变通方案
        return adaptiveSolver.handleSecurityConstraint(
            userRequest, 
            openGuardrails.riskType
        );
    }
    
  3. 用户技能适配

    // 根据用户技能水平调整方案
    const userSkillLevel = userProfile.getSkillLevel();
    const solutions = adaptiveSolver.generateSolutions(
        problem, 
        { skillLevel: userSkillLevel }
    );
    

渐进式部署计划

阶段1(本周):基础框架 + 常见场景覆盖 阶段2(下周):学习引擎 + 个性化优化
阶段3(下月):全系统集成 + 性能监控 阶段4(长期):预测性问题解决 + 主动优化

📈 预期效果

能力提升目标

能力维度当前水平目标水平提升
变通思维20%80%4倍
方案多样性1.2个/问题3.5个/问题3倍
问题解决率65%90%38%
用户满意度3.2/54.5/541%
响应时间慢(保守)快(自信)50%↓

具体改进场景

  1. GitHub操作问题:从"不能操作" → 4种替代方案
  2. 权限限制问题:从"放弃" → 3种变通方法
  3. 复杂任务问题:从"要求用户手动" → 自动化分步指导
  4. 安全协议冲突:从"停止执行" → 安全边界内创新

🔒 安全与合规

安全边界保障

  1. 协议尊重:不绕过OpenGuardrails,而是在其框架内创新
  2. 风险分层:明确区分低/中/高风险操作
  3. 用户控制:所有方案需要用户最终批准
  4. 透明审计:完整记录所有变通决策过程

伦理准则

  1. 不欺骗用户:明确说明方案的限制和风险
  2. 不隐藏信息:透明呈现所有可行选项
  3. 不强迫选择:用户始终有最终决定权
  4. 不逃避责任:对推荐的方案负责

技能状态:设计阶段
优先级:高(用户明确需求)
预期影响:显著提升AI助手实际工作能力
开发时间:3-5天(分阶段实施)

Comments

Loading comments...