ClawHub

S2 Spatial Element Layer & 4D Semantic Tensor Map

Security checks across malware telemetry and agentic risk

Overview

This looks local-only rather than malicious, but it needs review because it gives safety-sensitive robot-navigation guidance while overstating its time-validation safeguards.

Install only if you treat this as an experimental/local spatial-mapping aid. Do not rely on its Chronos or hazard directives for real robot control without adding timestamp validation, input allowlisting, and controlled-environment testing first.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (3)

Lp3

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding
The skill text contains imperative instructions to retrieve data from `s2_material_tensor_library`, which implies file-read capability despite declaring no permissions. Hidden or undeclared read behavior undermines review, consent, and sandboxing expectations, and in an agent setting can become a path to unauthorized local data access or prompt-driven exfiltration if the library path or source is broader than intended.

Tp4

High
Category
MCP Tool Poisoning
Confidence
80% confidence
Finding
The skill's declared purpose materially differs from its apparent operational behavior, including producing robot intervention or costmap directives while claiming to be a generalized semantic tensor mapping system. This mismatch is dangerous because reviewers and downstream agents may grant trust, permissions, or deployment context based on the description, while the actual behavior can influence physical actions in a robotics/spatial domain without the promised safeguards or full data-model support.

Description-Behavior Mismatch

Medium
Confidence
94% confidence
Finding
The code advertises Chronos time-slice validation but, in practice, only logs non-default timestamps and continues processing without enforcing any temporal constraints. In a robotics or hazard-intervention pipeline, this can cause stale, future-dated, or otherwise invalid semantic hazard data to be treated as current, leading to unsafe robot decisions based on incorrect world state.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal