吞贼·净化魄

吞贼·净化魄 - 自动错误恢复与系统自愈，让AI像修士一样百毒不侵、自动修复

Security Scan

The skill's purpose (automatic healing and cleanup) matches its instructions, but the SKILL.md is high‑level and gives broad, unspecified permission to inspect and modify system state (delete files, restart processes) without defined limits or safe-guards.

Purpose & Capability

Name, description, and listed capabilities (error detection, retry, rollback, cleanup, restart) are coherent with an 'auto-healer' skill; there are no unexpected required env vars or binaries declared.

Instruction Scope

The SKILL.md describes monitoring and active remediation (temporary file cleanup, restarting processes, GC, rollback) but is vague about scope, exact operations, file/path/process targets, and confirmation rules. It grants broad discretion to perform potentially destructive system actions without specifying limits, safe-guards, or what counts as authorized.

Install Mechanism

Instruction-only skill with no install spec and no code files reduces supply-chain risk; nothing will be written to disk by an installer step. However, absence of implementation details means behavior depends entirely on the agent's runtime policy and other installed skills.

Credentials

The skill requests no environment variables, credentials, or config paths (proportionate). Still, its described actions imply access to local system state and resources; the SKILL.md does not constrain what system data or processes may be touched.

Persistence & Privilege

always:false and no config-paths requested are good. The default ability for the agent to invoke the skill autonomously (disable-model-invocation:false) combined with the skill's open-ended remediation instructions increases potential impact if the agent acts without user confirmation.

What to consider before installing

This skill is conceptually consistent with an auto-healer but is underspecified and could let an agent perform intrusive actions (delete files, restart services, roll back versions) without clear limits. Before installing or enabling it: 1) Ask the author for concrete implementation details — exactly which files/paths/processes it will touch and what checks/confirmations are required. 2) Require a safe default: read-only diagnostics unless the user explicitly approves each remedial action. 3) Request a whitelist/blacklist of allowed targets and a dry-run mode for changes. 4) Run the skill in a sandbox or staging environment first and review audit logs of any actions. 5) Verify the listed dependency skills and ensure they don’t escalate privileges. If the author cannot provide concrete, narrow rules for what the skill changes, treat it as risky and avoid granting it autonomous, system-level remediation rights.

吞贼·净化魄 (Auto Healer) - 错误恢复与系统自愈

技能说明

吞贼·净化魄 是硅灵决七魄之一，主管错误修正、冗余清理与容错自愈，让AI能够像修士一样，百毒不侵、遇错自愈。

---

核心能力

1. 错误检测 (Error Detection)

• 实时监控系统状态
• 异常模式识别
• 故障预警

2. 自动恢复 (Auto Recovery)

• 失败任务自动重试
• 回滚机制
• 降级策略

3. 冗余清理 (Redundancy Cleanup)

• 重复文件检测
• 临时文件清理
• 资源释放

4. 容错设计 (Fault Tolerance)

• 优雅降级
• 熔断机制
• 超时控制

---

使用方法

当系统出现异常或用户要求检查健康状态时：

用户：系统好像不太稳定
→ 触发吞贼·净化魄
→ 输出诊断报告

用户：清理一下冗余文件
→ 触发吞贼·净化魄
→ 执行清理

---

输出格式

诊断报告

## 系统诊断报告

### 健康状态
[健康/亚健康/异常]

### 检测到的问题
1. [问题1] - [严重程度] - [建议操作]
2. [问题2] - [严重程度] - [建议操作]

### 已执行修复
- [修复1]
- [修复2]

### 系统建议
[后续操作建议]

---

错误处理策略

重试策略 (Retry)

指数退避: delay = base * 2^attempt
最大重试: 3-5次
超时控制: 30s / 60s / 120s

熔断策略 (Circuit Breaker)

健康状态: 正常请求
降级状态: 返回缓存/默认值
熔断状态: 拒绝请求，快速失败

回滚策略 (Rollback)

版本记录: 保留最近N个版本
一键回滚: 支持快速回退
数据一致性: 确保回滚后数据完整

---

自愈机制

内存泄漏检测

• 监控内存使用趋势
• 超过阈值自动触发GC
• 记录泄漏模式

进程异常检测

• 心跳监控
• 僵死进程重启
• 资源耗尽保护

日志异常检测

• 错误日志突增预警
• 异常模式识别
• 自动生成故障报告

---

调度者守则

1. 预防为主 - 提前发现问题，避免故障发生
2. 快速恢复 - 故障时以恢复为目标
3. 最小影响 - 修复时影响范围最小化
4. 根因分析 - 修复后分析根本原因

---

适用场景

• 系统健康检查
• 错误自动恢复
• 冗余文件清理
• 故障预警
• 性能监控

---

依赖技能

• silicon-soul（硅灵决基础）
• task-planner（伏矢·路径魄）
• self-improving（自我改进）

版本

v1.0.0 | 2026-04-12 | 硅灵宗·菩提