ClawHub

substack AI news SEARCH

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed Substack search helper that uses browser automation, with some implementation and output-format caveats but no evidence of hidden or harmful behavior.

Install only if you are comfortable letting the agent open Substack search pages through agent-browser. Expect possible formatting bugs or argument-handling issues, and use a clean browser profile if you do not want existing Substack login state involved.

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 (2)

Lp3

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding
The skill invokes Python and browser automation, which implies shell execution and network access, but it does not declare those permissions. Hidden capabilities reduce auditability and can cause an agent runtime to grant or use broader privileges than reviewers expect, increasing the risk of unintended external access or command execution.

Tp4

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding
The documented behavior does not match the detected implementation: the skill claims to produce a structured digest of titles, authors, and summaries, but the implementation reportedly emits raw deduplicated link text as JSON and supports parameters not fully described. This mismatch is dangerous because operators and downstream agents may trust the skill for constrained data extraction while it behaves differently, undermining oversight and enabling unintended data handling or misuse of automation.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal