ClawHub
Back to skill

Security audit

Audit Log Hook

Security checks across malware telemetry and agentic risk

Overview

This audit skill is not clearly malicious, but it would broadly log tool activity and sensitive context while giving incomplete or misleading protection details.

Install or implement this only if you intentionally want broad audit logging of agent tool activity. Before use, change it to log metadata by default, apply real recursive redaction before any logging, avoid raw tool results unless explicitly needed, restrict log access, and define retention and cleanup.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Intent-Code Divergence

High
Confidence
97% confidence
Finding
The skill claims sensitive fields are auto-redacted, but the actual hook examples log raw tool parameters and results and never call the redaction helper. This creates a mismatch between documented safety guarantees and real behavior, making secrets, credentials, personal data, and command output likely to be exposed in logs.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill is designed to collect tool calls, session identifiers, and user identifiers for auditing, but it provides no notice, consent flow, retention policy, or privacy controls. In a system that may process sensitive prompts, commands, or user-linked activity, this creates an unnecessary privacy exposure and compliance risk.

Missing User Warnings

High
Confidence
96% confidence
Finding
The example hook code records raw tool parameters before execution and raw results after execution, including session and user identifiers, without applying redaction or warning about sensitive capture. Because tool params and outputs commonly contain tokens, file contents, command output, and personal data, this materially increases the chance of sensitive-data leakage through logs.

Ssd 3

Medium
Confidence
84% confidence
Finding
The skill explicitly promotes storing user/tool inputs and outputs in plain-language audit records, which creates a direct collection and exposure path for sensitive natural-language content. Even if intended for debugging or statistics, such logs can accumulate secrets, private conversations, or regulated data and become a high-value target.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.