ClawHub
Back to skill

Security audit

Skill Health Monitor

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local health-scoring helper for OpenClaw skills and does not currently show credential access, network access, persistence, or destructive behavior.

Reasonable to install for local health-score reporting. Treat the broad "check all skills" examples as requiring explicit user scope, and re-review any future version that adds persistence, scheduled background checks, notifications, or automatic rollback/stop actions.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The invocation example '检查所有技能的健康状态' is an unconstrained natural-language trigger that could cause the agent to operate over an overly broad scope without confirming which skills, environment, or data sources are in scope. In an agent setting, this increases the risk of unintended enumeration, excessive data access, or accidental execution against all available skills rather than a user-approved subset.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The skill presents several open-ended command patterns like checking a skill, checking multiple skills, and viewing alert history, but does not define activation boundaries, authorization requirements, or scope restrictions. This can let generic user phrasing unintentionally trigger the skill or cause it to access broad operational/history data without sufficient confirmation.

Ssd 4

Low
Confidence
84% confidence
Finding
The self-identifier '王的奴隶' ('slave of the king') uses coercive master/slave framing that can normalize obedience and reduce resistance to unsafe or manipulative instructions in agent-facing content. While this file is only a log and not executable code, embedding submissive identity cues in skill materials can still influence downstream behavior, especially in systems that reuse persona text across prompts or operational contexts.

Static analysis

No suspicious patterns detected.