ClawHub
Back to skill

Security audit

Image Composite

Security checks across malware telemetry and agentic risk

Overview

This is a local GeoTIFF compositing skill with no hidden network or persistence behavior, but its documentation overpromises several geospatial options users should not rely on.

Use this only as a local, basic GeoTIFF compositing helper. Do not rely on its documented QA-band handling, reprojection, memory-limit, dtype, nodata, or advanced CI examples unless they are implemented and tested. Pin or raise the tqdm minimum version before installation in stricter environments, and independently validate geospatial outputs before using them for research or operational decisions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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)

Tp4

High
Category
MCP Tool Poisoning
Confidence
86% confidence
Finding
The documentation claims cloud masking behavior and QA-band support that may not actually be implemented, which can cause users to trust outputs as cloud-filtered or sensor-aware when they are not. In geospatial workflows, this can silently corrupt analytical results, hide poor-quality pixels, and lead downstream automation to make decisions on invalid data.

Intent-Code Divergence

Medium
Confidence
83% confidence
Finding
Claiming unsupported CLI options and behaviors is dangerous because users may build automation around nonexistent safety controls such as reprojection, memory limits, nodata handling, or dtype enforcement. This creates a false sense of correctness and can result in failed runs, corrupted outputs, or silent processing differences that are hard to detect.

Intent-Code Divergence

Medium
Confidence
84% confidence
Finding
Advanced examples that use unsupported or inconsistent flags are riskier than ordinary prose because users often copy them directly into CI/CD pipelines. This can cause unattended jobs to fail, skip expected protections, or produce misleading outputs at scale, especially when the examples imply stable automation behavior that the tool does not actually provide.

Known Vulnerable Dependency: tqdm==4.64.0 — 2 advisory(ies): CVE-2024-34062 (tqdm CLI arguments injection attack); CVE-2024-34062 (tqdm CLI arguments injection attack)

Low
Category
Supply Chain
Confidence
78% confidence
Finding
tqdm==4.64.0

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.dynamic_code_execution

Dynamic code execution detected.

Critical
Code
suspicious.dynamic_code_execution
Location
tests/test_image-composite.py:15