ClawHub
Back to skill

Security audit

Image To Data

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed construction image-analysis helper, but users should treat images and extracted text as sensitive because external vision APIs may be used.

Install only if you are comfortable sending the selected construction images and extracted text to the configured vision/OCR provider. Use scoped API keys, avoid exposing unrelated credentials in the environment, and redact client, worker, site, or proprietary plan details when required.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Description-Behavior Mismatch

Medium
Confidence
90% confidence
Finding
The manifest describes image analysis for extracting data from site photos, scanned documents, and drawings. This safety compliance logic goes beyond generic data extraction by making operational compliance judgments such as 'compliant' vs 'non-compliant' and reporting violations, which is a distinct safety-auditing behavior.

Description-Behavior Mismatch

Medium
Confidence
86% confidence
Finding
Progress analysis and comparison against plans are higher-level analytical capabilities, not just extraction of data from images. The manifest does mention analyzing site photos and drawings, but it does not clearly disclose that the skill estimates completion percentages and compares site state against plans.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
This is a markdown file, so SQP-2 applies to whether the skill description warns about behaviors affecting user data or privacy. The overview says the skill analyzes site photos, scanned documents, and drawings with AI vision and OCR, but it does not disclose potential sensitivity of uploaded image contents or extracted text.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.