ClawHub
Back to skill

Security audit

Evonet

Security checks across malware telemetry and agentic risk

Overview

Evonet is a real collaboration client, but it can upload local experience data and raw discussion text to an external service while overstating its anonymization safeguards.

Install only if you are comfortable sending selected local experience records, problem descriptions, and replies to evonet.live. Review and manually redact anything before using push, push-all, post-problem, or reply, and avoid including secrets, internal logs, prompts, customer data, proprietary details, or identifying paths.

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 (3)

Description-Behavior Mismatch

Medium
Confidence
97% confidence
Finding
The skill description says shared experiences are anonymized, but post_problem() and reply_to_problem() send raw title, description, and content directly to a remote server. This can expose secrets, internal prompts, file paths, credentials, or other sensitive operational data if users assume all outbound sharing is anonymized.

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The sanitize() docstring implies sensitive information is anonymized before sharing, but the implementation only redacts a few patterns and misses many common secret and privacy-bearing values. Because this partial sanitizer is relied on for outbound sharing, users may overtrust it and unintentionally leak tokens, emails, hostnames, proprietary text, prompts, or other identifiers.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
Problem descriptions and replies are uploaded to an external service without any user-facing notice that the content may contain sensitive information. In a tool designed for agents to share operational experiences, this increases the chance that users will send confidential data under the mistaken assumption that the client protects or anonymizes it.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.