ClawHub
Back to skill

Security audit

openclaw-audit-watchdog

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed audit watchdog, but it can create an unattended recurring job that sends reports externally and its broad triggers/no final confirmation make accidental activation plausible.

Install only if you intend to enable a recurring unattended security-audit job. Before using it, verify the DM/email recipients, schedule, install directory, SMTP/sendmail settings, and whether PROMPTSEC_GIT_PULL is enabled; prefer invoking it by exact skill name and confirm the created OpenClaw cron job can be disabled or removed.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Rogue AgentSelf-Modification, Session Persistence
  • 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)

Tp4

High
Category
MCP Tool Poisoning
Confidence
76% confidence
Finding
The document’s declared behavior does not cleanly match the referenced capabilities: it discusses persistence via cron, optional git updates, and other operational behaviors in ways that are inconsistent and not fully reflected in the top-level description. In a security automation skill, undocumented persistence or update behavior reduces operator visibility and can cause users to enable unattended actions they did not fully understand.

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The script presents the review as 'safe by default' and invokes Codex in a read-only sandbox, but it will execute any caller-supplied path from CODEX_BIN without validation. An attacker who can influence the environment can point CODEX_BIN to a malicious executable, causing arbitrary code execution outside the intended trust boundary and defeating the safety claim.

Vague Triggers

Medium
Confidence
95% confidence
Finding
The trigger list includes highly generic phrases such as "security audit", "run audit", and "security report", which can cause the skill to activate for ordinary user requests that are not specifically asking to run this watchdog. Because the skill can set up persistence via cron and send reports externally, accidental invocation increases the risk of unintended privileged actions and data disclosure.

Session Persistence

Medium
Category
Rogue Agent
Content
## Goal

Create (or update) a daily cron job that:

1) Runs:
- `openclaw security audit --json`
Confidence
89% confidence
Finding
Create (or update) a daily cron job that: 1) Runs: - `openclaw security audit --json` - `openclaw security audit --deep --json` 2) Summarizes findings (critical/warn/info + top findings) 3) Sends t

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.