ClawHub
Back to skill

Security audit

Screenshot Visual

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed web reconnaissance screenshot skill; it has legal and data-handling risks if used outside authorized scope, but the reviewed artifacts do not show deception, exfiltration, persistence hooks, or destructive behavior.

Install only if you intend to run authorized web reconnaissance, not just capture one exact page. Keep the generated screenshots, JSON files, and reports private, and delete them when no longer needed because they may contain sensitive page content or exposed input names.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The skill is described as screenshot capture and visual/contextual analysis, but it actively probes a hard-coded list of sensitive paths such as admin, debug, config, and API endpoints. That expands behavior from passive capture into targeted discovery, which can surprise users, increase legal/operational risk, and cause the agent to access resources beyond what the stated capability implies.

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The code deliberately checks known sensitive locations like /admin, /.env, /phpmyadmin, /debug, and /swagger, which is more akin to reconnaissance than simple screenshotting. In an agent skill, this creates a capability mismatch and may facilitate unauthorized discovery of exposed administrative or secret-bearing resources.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill makes requests to multiple additional endpoints and captures screenshots and metadata without clearly warning the user that it will broaden scope beyond the provided URL. This can lead to unintended collection from sensitive pages and creates privacy, compliance, and authorization risks, especially when used against third-party systems.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The script stores screenshots, page snippets, headers, form metadata, and potentially sensitive input identifiers to local disk without explicit notice or data minimization. If scanned pages contain secrets, internal admin panels, tokens, or personal data, this creates a secondary exposure on the operator's filesystem and in generated reports.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.