ClawHub
Back to skill

Security audit

design-analysis

Security checks across malware telemetry and agentic risk

Overview

This skill is a local design-report generator that scans a user-chosen image folder and writes a user-chosen HTML file, with some logging and overwrite caveats but no evidence of hidden exfiltration or destructive behavior.

Install only if you are comfortable letting the skill read images from the folder you specify and create or replace the HTML file path you provide. Use a dedicated design-assets folder, choose a new output filename, and avoid passing secrets in custom sections or context because the wrapper logs full parameters.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (3)

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
The test suite builds a shell command by concatenating unquoted values and passes it to execSync, which invokes a shell. Although the current test cases are hardcoded, values such as input, output, title, and skillPath could become dangerous if later made user-controlled or if paths contain shell metacharacters, leading to command injection or unintended command execution.

Missing User Warnings

Low
Confidence
95% confidence
Finding
The skill asks users to provide a local input folder and an output file path, but it does not clearly warn that it will scan local files and write a generated HTML artifact to disk. This can lead to unintended disclosure or modification of local data, especially if users do not realize the scope of filesystem access implied by the workflow.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The wrapper logs the full params object, which may include sensitive filesystem paths, user-supplied data, and the OpenClaw context object. In agent environments, context often carries tokens, identifiers, prompts, or other confidential metadata, so verbose logging can leak secrets to logs, consoles, or downstream log aggregation systems.

VirusTotal

47/47 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.dangerous_exec

Shell command execution detected (child_process).

Critical
Code
suspicious.dangerous_exec
Location
test.js:60