ClawHub
Back to skill

Security audit

Agent安全审计免费版

Security checks across malware telemetry and agentic risk

Overview

This skill is a local Agent security-audit helper whose sensitive file access is mostly expected for its stated purpose, with a few documentation ambiguities users should notice.

Install only if you want an agent to run local grep/find/bash-style checks over a target Agent project, including configuration and .env files. Review any callback or export use before enabling it, and keep scan output local if it may mention sensitive filenames, secret patterns, or configuration weaknesses.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (4)

Description-Behavior Mismatch

Medium
Confidence
80% confidence
Finding
The skill presents itself as a local/free audit utility, but its documented input includes a callback_url and export/query-style behavior that can enable outbound communication not clearly bounded by the core local-audit promise. This creates a trust and data-flow mismatch: users may supply sensitive code or findings expecting only local processing, while the interface suggests optional remote notification or integration paths.

Intent-Code Divergence

Medium
Confidence
68% confidence
Finding
The free-edition comparison says continuous monitoring is not supported, but other sections imply query/export or network-assisted operation. While this is not direct code execution, inconsistent operational claims can mislead users about persistence, telemetry, and attack surface, which is a security-relevant documentation flaw.

Vague Triggers

Medium
Confidence
71% confidence
Finding
The capability coverage text defines an extremely broad and ambiguous trigger scope, spanning generic security-audit and agent-related terms. In an agent environment, overbroad invocation can cause the skill to activate on unrelated contexts and unnecessarily gain access to read/exec/write workflows, increasing the chance of unintended execution or data exposure.

Credential Access

High
Category
Privilege Escalation
Content
echo "=== Agent配置安全审计 ==="
# ...
# 检查配置文件
for config in config.json config.yaml .env agent_config.json settings.json; do
    if [ -f "$config" ]; then
        echo ""
        echo "--- 检查: ${config} ---"
Confidence
83% confidence
Finding
.env

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.prompt_injection_instructions

Prompt-injection style instruction pattern detected.

Warn
Code
suspicious.prompt_injection_instructions
Location
SKILL.md:426