ClawHub
Back to skill

Security audit

Self-Hosted Funnel Launch

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent deployment and operations guide for a self-hosted funnel builder, with disclosed credential, payment, hosting, and MCP write-access considerations.

Before installing, treat any write-enabled MCP API key as production authority over your funnel tenant. Use read-only keys unless the agent must edit, create one key per agent, revoke instead of sharing, keep secrets out of config files, verify drafts before publishing, and confirm the documented page-reuse and thank-you/error-page behavior against the current Autonnel version.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Intent-Code Divergence

Medium
Confidence
98% confidence
Finding
L184 states that a `LANDING` page can belong to only one funnel and that reuse is rejected, but later guidance at L238 says the opposite: the same page may be referenced by several funnels and there is no one-funnel-only rule. This is an active contradiction in the skill's operational guidance and can cause an agent to make incorrect assumptions about valid write operations.

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
L185 says thank-you and error pages are auto-bound only from existing pages and must be created first if none exist, but L240 says `create_funnel` creates them from templates if none exist. These statements cannot both be true and create conflicting operator/agent expectations around funnel creation behavior.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.