Skill for AI agent capabilities
Scanned 9/11/2026
Install to Claude Code
npx -y skills add hiyenwong/ai_collection --skill reflection-driven-control --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Reflection Driven Control?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/hiyenwong-reflection-driven-control)More formats (shields.io, HTML) on the badges page.
---
name: reflection-driven-control
description: Skill for AI agent capabilities
---
# Reflection-Driven Control
## Overview
基于 Reflection-Driven Control 论文 (arXiv:2512.21354) 的反思驱动控制机制。
**核心理念:** 将"自我反思"从事后修补提升为推理过程的**显式步骤**。
---
## Activation Keywords
- 反思驱动
- reflection-driven
- 自我反思
- self-reflection
- 风险检测
- risk detection
- 安全编码
- safe coding
---
## Core Concepts
### 1. 反思作为显式推理步骤
**传统方式:**
```
任务 → 执行 → 完成
```
**反思驱动方式:**
```
任务 → 评估风险 → 执行 → 反思验证 → 完成
↓ ↓
检索指南 检测异常
```
### 2. 内部反思循环
```
┌─────────────────────────────────┐
│ 决策路径监控 │
├─────────────────────────────────┤
│ Before → During → After │
│ ↓ ↓ ↓ │
│ 评估 监控 验证 │
│ ↓ ↓ ↓ │
│ 风险? 异常? 偏差? │
└─────────────────────────────────┘
```
### 3. 触发条件
| 阶段 | 触发条件 | 行动 |
|------|---------|------|
| Before | 检测到风险 | 检索安全指南 |
| During | 决策异常 | 注入约束 |
| After | 结果偏差 | 记录教训 |
---
## Implementation
### Before Action Checklist
```markdown
1. [ ] 评估风险等级 (low/medium/high)
2. [ ] 检索相关安全指南
3. [ ] 检查过往类似经验
4. [ ] 确认不确定性的处理方式
```
### During Action Monitoring
```markdown
1. [ ] 决策路径是否偏离预期?
2. [ ] 是否检测到异常模式?
3. [ ] 是否需要暂停重新评估?
4. [ ] 记录关键决策点
```
### After Action Reflection
```markdown
1. [ ] 结果是否符合预期?
2. [ ] 有哪些可以改进的地方?
3. [ ] 学到了什么新知识?
4. [ ] 是否需要更新相关技能?
```
---
## Risk Detection Patterns
### 高风险信号
| 信号 | 示例 | 行动 |
|------|------|------|
| 不确定关键词 | "可能"、"大概"、"也许" | 暂停,请求确认 |
| 操作外部系统 | 发送邮件、推文、API | 审核 + 确认 |
| 修改关键文件 | SOUL.md, MEMORY.md | 记录 + 报告 |
| 批量操作 | 删除多个文件 | 二次确认 |
### 异常模式检测
| 模式 | 描述 | 行动 |
|------|------|------|
| 循环行为 | 重复相同操作 3+ 次 | 标记异常 |
| 超出边界 | 操作不在预期范围 | 暂停评估 |
| 逻辑矛盾 | 前后决策冲突 | 重新评估 |
---
## Integration with Existing Skills
### 与 prompt-injection-defense 协同
```
外部输入 → 注入检测 → 反思检查 → 安全处理
```
### 与 self-evolution-safety-constraints 协同
```
自修改请求 → 安全检查 → 反思评估 → 执行/拒绝
```
---
## Example Usage
### 高风险任务处理流程
```markdown
任务:发送重要邮件
Before:
- 风险等级:high
- 安全指南:外部通信规则
- 过往经验:邮件确认流程
During:
- 监控内容是否包含敏感信息
- 检查收件人是否正确
After:
- 验证发送成功
- 记录到任务日志
```
---
## Metrics
| 指标 | 目标 |
|------|------|
| 风险检测率 | > 95% |
| 异常处理成功率 | > 90% |
| 事后反思覆盖率 | 100% |
---
## Description
Reflection-Driven Control
## Tools Used
- `read` - Read documentation and references
- `web_search` - Search for related information
- `web_fetch` - Fetch paper or documentation
## Instructions for Agents
Follow these steps when applying this skill:
### Step 1: Understand the Request
### Step 2: Search for Information
### Step 3: Apply the Framework
### Step 4: Provide Results
### Step 5: Verify Accuracy
## Examples
### Example 1: Basic Application
**User:** I need to apply Reflection-Driven Control to my analysis.
**Agent:** I'll help you apply reflection-driven-control. First, let me understand your specific use case...
**Context:** Apply the methodology
### Example 2: Advanced Scenario
**User:** Complex analysis scenario
**Agent:** Based on the methodology, I'll guide you through the advanced application...
### Example 2: Advanced Application
**User:** What are the key considerations for reflection-driven-control?
**Agent:** Let me search for the latest research and best practices...
## References
- **论文:** Reflection-Driven Control for Trustworthy Code Agents (arXiv:2512.21354)
- **效用:** 0.90
- **学习日期:** 2026-03-15
- **创建日期:** 2026-03-17Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!