ClawHub
Back to skill

Security audit

rt-attention-analysis

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local reaction-time data analysis helper that reads user-provided datasets and writes reports without hidden network, credential, or persistence behavior.

Before installing, users should understand that the skill runs local Python scripts over their RT datasets and can overwrite files at the output paths they choose. Use a dedicated output directory for reports if preserving prior results matters.

SkillSpector

Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (1)

Lp3

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding
Without declared permissions the skill's intent is opaque and cannot be validated.

Static analysis

No suspicious patterns detected.