ClawHub

Agent Oversight

Security checks across malware telemetry and agentic risk

Overview

This looks like a small local oversight helper, but it overstates safety features and describes automatic session killing without safeguards.

Review before installing. Treat this as an unfinished oversight helper: it can list local OpenClaw sessions, but it should not be relied on for real coordination, failure logging, or safe hung-session cleanup unless those behaviors are implemented with explicit confirmation and clear limits.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (2)

Lp3

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding
The skill advertises executable shell-based behavior via `python3 oversight.py ...` and operational capabilities like killing sessions, but it declares no permissions. This creates a trust and review gap: consumers may invoke a skill with process-management and file-writing effects without explicit authorization boundaries or visibility into what it can access or modify.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
Automatically killing 'hung' sessions after a fixed threshold without warning or confirmation can terminate legitimate long-running work, causing data loss, denial of service, or interruption of other agents. In an oversight skill with authority over sub-agents, this is more dangerous because it centralizes the ability to disrupt multiple active tasks based on potentially inaccurate heuristics.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal