ClawHub
Back to skill

Security audit

Rules of the Claw

Security checks for vulnerabilities and agentic risk

Overview

This is a protective Guardian ruleset that persistently replaces the local Guardian rule file; its broad blocking and placeholders need review, but no hidden exfiltration or destructive behavior was found.

Install this only if you want Guardian to enforce a broad persistent security baseline. Review the JSON first, keep the installer backup, customize the placeholders for your app/org/user, and disable rules that are too noisy for your workflow.

SkillSpector

Vulnerability Patterns
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • 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
Findings (2)

Vague Triggers

Medium
Confidence
76% confidence
Finding
Using an undefined approval term like 'YOUR_USER' creates policy ambiguity around who can authorize dangerous operations. In agent safety systems, unclear approval semantics can lead to inconsistent enforcement, social-engineering opportunities, or unsafe operator assumptions about when destructive actions are permitted.

Tool Parameter Abuse

High
Category
Tool Misuse
Content
"tool": "exec",
    "pattern": "rm\\s+-(r|rf|fr)\\s+.*(workspace|your_app)",
    "field": "command",
    "blockMessage": "🛡️ rm -rf blocked on workspace/your_app. Use trash or ask YOUR_USER.",
    "severity": "high"
  },
  {
Confidence
75% confidence
Finding
The block message tells operators to 'ask YOUR_USER,' but that approval authority is undefined and could be socially engineered or interpreted inconsistently. For destructive commands like rm -rf, ambiguous escalation guidance can undermine the control's effectiveness and create a procedural bypass.

Static analysis

Detected: suspicious.prompt_injection_instructions

Prompt-injection style instruction pattern detected.

Warn
Code
suspicious.prompt_injection_instructions
Location
SKILL.md:33