Back to skill

Security audit

agent-desktop-windows

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because its troubleshooting guidance raises system-wide security and data-loss concerns; the scan was partial because one local reference could not be resolved.

Review the troubleshooting guidance before installing. Consider whether you are comfortable with advice that weakens a system-wide security control and an example that could overwrite unsaved text. The scanners inspected all five listed components, but coverage is still marked partial because a local reference could not be resolved, so the evidence is incomplete.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

The troubleshooting instructions recommend disabling Smart App Control without warning about the system-wide security impact.

High
Category
Output Handling
Confidence
99% confidence
Finding

The guide recommends turning off Smart App Control to run an unsigned executable without explaining that this weakens application-execution protection system-wide, not just for agent-desktop. In a desktop automation skill, an agent could interpret this troubleshooting step as authorization to disable a security control, allowing other untrusted executables to run; offering a signed build as an alternative does not adequately constrain that risk.

Content

No source excerpt is available for this finding.

The save demonstration replaces the contents of an existing Notepad document without warning about loss of unsaved text.

Medium
Category
Output Handling
Confidence
96% confidence
Finding

The demonstration explicitly targets an already-open Notepad and replaces the selected document field with fixed demo text before saving elsewhere. Because it neither creates a fresh document nor preserves existing contents, an agent following it can overwrite a user's unsaved work or target the wrong document when multiple Notepad instances are open; this is an unsafe automation example rather than evidence of malicious intent.

Content

No source excerpt is available for this finding.