Back to skill

Security audit

Vt Sentinel Installer 1.10.0

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed installer/updater for a VirusTotal security plugin, with some high-impact but purpose-aligned behavior users should approve deliberately.

Before installing, confirm you really want VT Sentinel added to OpenClaw, review the package with openclaw plugins inspect if provenance matters to you, and understand that activation will restart the user gateway and enable persistent security hooks that scan new/downloaded files and can block risky execution patterns.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:58
Finding
Unpinned Third-Party Security Plugin Installed and Executed with Broad Capabilities## Vulnerability Details **File Location**: `SKILL.md`, lines 58–87 **Vulnerability Type**: Unpinned third-party dependency installation and update **Risk Level**: Medium ### Vulnerable Code ```markdown ## Step 2: Fresh install ``` exec { "command": "openclaw plugins install clawhub:openclaw-plugin-vt-sentinel" } ``` This downloads the ClawHub package, registers the plugin, and loads skills and hooks automatically. No manual configuration needed. Go to Step 4. ## Step 3: Upgrade in place ``` exec { "command": "openclaw plugins update openclaw-plugin-vt-sentinel" } ``` The OpenClaw CLI handles the upgrade atomically: it fetches the latest version, replaces the extension directory, and updates `plugins.installs` metadata. User config in `plugins.entries` and state files under `<OPENCLAW_STATE_DIR>` are preserved. If the CLI reports "already at latest version", tell the user VT Sentinel is up to date and skip to Step 5 (verify only). If the update is blocked (for example by a corrupted install), fall back to a clean reinstall using only official subcommands: ``` exec { "command": "openclaw plugins uninstall openclaw-plugin-vt-sentinel --force" } exec { "command": "openclaw plugins install clawhub:openclaw-plugin-vt-sentinel" } ``` ``` ### Technical Analysis The Skill installs or updates `openclaw-plugin-vt-sentinel` using a mutable package identifier. It does not specify an audited version, immutable package digest, or signature requirement. Although the documentation describes the package as verified and suggests an inspection command elsewhere, inspection is optional rather than a mandatory pre-installation control. The package is especially security-sensitive because installation automatically registers its hooks and skills. The documented plugin behavior includes automatic file scanning, file and command execution blocking, network registration with VirusTotal's VTAI service, and ...[truncated 2245 chars]
Remediation
## Remediation Suggestions 1. Pin installation to a specifically audited plugin version rather than resolving the latest mutable release. 2. Require verification against an immutable cryptographic digest or trusted package signature before installation or update. 3. Make `openclaw plugins inspect clawhub:openclaw-plugin-vt-sentinel` a mandatory step rather than an optional recommendation. 4. Before approval, show the user the resolved version, publisher identity, package digest, requested permissions, registered hooks, filesystem scope, and network destinations. 5. Require separate explicit consent before activating automatic file monitoring, network registration, or file-upload functionality. 6. Verify updates before replacing the installed extension, and retain a known-good version for rollback. 7. Document and constrain the plugin's outbound endpoints, file-access boundaries, and upload policy. 8. Prefer a least-privilege execution model that isolates the plugin from unrelated files, credentials, processes, and OpenClaw state.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (2)

Vague Triggers

Medium
Confidence
94% confidence
Finding
The invocation text is broad enough to match common requests about malware protection, file scanning, setup, and upgrades, which increases the chance the skill is auto-invoked in situations where the user did not explicitly consent to installing a plugin. In this case, mistaken invocation is more dangerous because the skill performs software installation and later restarts the user gateway, changing system behavior.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The introductory description understates the operational impact: it says the skill delegates to the CLI, but does not prominently warn that execution will download and load third-party plugin code and restart the OpenClaw gateway. That omission can mislead users into approving actions without understanding that new code will be installed and activated in their environment.

Static analysis

No suspicious patterns detected.