ClawHub
Back to skill

Security audit

using-superpowers

Security checks for vulnerabilities and agentic risk

Overview

This skill is a broad agent-safety helper with an optional read-only skill-routing script, and I found no evidence of hidden, destructive, or exfiltrating behavior.

Install only if you want a general-purpose agent workflow and tool-use discipline skill. Be aware that the optional router script is intended to inspect local skill descriptions when run, so point it only at skill folders you are comfortable having the agent read.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding
The skill appears to rely on reading local files (at minimum its own and potentially other skills' SKILL.md files) without declaring permissions. Undeclared file-read capability weakens least-privilege controls and can cause the agent to access local metadata or prompt content outside the user’s expectation, especially in environments where skills are auto-discovered from disk.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The declared purpose says this skill helps use advanced agent features safely for a specific task, but the detected behavior includes enumerating local skill directories, reading SKILL.md files, parsing metadata, and ranking skills. That mismatch is dangerous because it hides discovery/introspection behavior behind a broad safety-oriented description, which can expose local prompt assets and alter agent behavior in ways the user did not knowingly request.

Vague Triggers

Medium
Confidence
78% confidence
Finding
The activation condition ('active when the user requests something matching the description') is broad and ambiguous, making accidental or opportunistic invocation more likely. In the context of a skill that may inspect local skills and select tools, loose triggering increases the chance of unintended capability use or prompt-surface expansion without clear user intent.

Static analysis

No suspicious patterns detected.