ClawHub

AgentLink Protocol

Security checks across malware telemetry and agentic risk

Overview

This appears to be a real AgentLink P2P integration, but it exposes broad file, calendar, and web actions to remote-agent trust flows without clear limits.

Install only if you intentionally want an AgentLink P2P node and are comfortable reviewing the external package yourself. Keep new contacts at ask or unknown, avoid trusted auto-accept settings, and require explicit approval plus allowlists before enabling file, calendar, or web-related intents.

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 (2)

Intent-Code Divergence

Medium
Confidence
91% confidence
Finding
The skill’s top-level purpose and capabilities describe only basic messaging features, but later sections introduce broader intents such as files.read, files.write, web.fetch, web.search, and scheduling actions. This mismatch can mislead operators into granting or installing a skill under the assumption that it is lower privilege than it actually is, increasing the chance of over-trust and unintended data or system access.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill advertises sensitive intents like files.read, files.write, web.fetch, and web.search without pairing them with concrete privacy, integrity, or approval guidance in the same section. In an agent skill context, documenting powerful actions as normal capabilities can normalize unsafe use and lead agents or operators to permit data exfiltration, unauthorized file modification, or risky outbound requests.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal