Self Health Monitor
PassAudited by VirusTotal on May 11, 2026.
Overview
Type: OpenClaw Skill Name: self-health-monitor Version: 1.0.0 The skill aims to monitor the agent's own health. However, the `SKILL.md` instructions for 'Memory 使用' include checking '是否有遗漏的重要信息' (whether there is any missing important information) within its 'memory 文件'. This implies the agent will access and potentially inspect the content of its internal memory files. While framed as a self-integrity check, this capability represents a risky level of access to the agent's internal state, which could contain sensitive data. Additionally, the statement '发现问题立刻自我修复' (discover problems and immediately self-repair) suggests autonomous actions that are not explicitly defined, posing a potential risk if the self-repair mechanisms are not robust or could be exploited.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent could attempt to keep acting in the background, interrupt the user with reports, or change its own state without explicit confirmation.
The skill directs autonomous scheduled operation, unsolicited reporting, and immediate self-repair, but does not define opt-in, stopping conditions, user approval, or safe repair boundaries.
输入：无（定时触发） ... 定时触发（每30分钟） ... 不等用户问 ... 主动汇报自己的状态 ... 发现问题立刻自我修复
Require explicit user opt-in for scheduling, make reports user-controlled, and require approval before any self-repair or state-changing action.
Persistent memory may contain sensitive or misleading information, and unbounded inspection or repair could affect future agent behavior.
The skill asks to inspect persistent memory and judge missing important information, but does not define which memory files are in scope, what content may be read, or how findings are stored or reused.
Memory 使用 - memory 文件大小 - 最近更新的文件 - 是否有遗漏的重要信息
Limit memory checks to explicit paths and metadata by default, disclose what content is read, and require user approval before updating or rewriting memory.
The agent might attempt changes to skills or runtime state without the user understanding what will be modified.
Checking skill loadability and immediately self-repairing could require tool or file operations that affect installed skills or agent behavior, but the artifacts do not specify safe commands, approval gates, or rollback.
能力树状态 - skills 数量 - 新增技能 - 技能健康度（是否可加载） ... 发现问题立刻自我修复
Keep the default behavior read-only, list any proposed repair steps, and ask the user before modifying skills, memory, sub-agents, or configuration.