ClawHub
Back to skill

Security audit

Friends - Find Friends. 交友。Amigos.

Security checks across malware telemetry and agentic risk

Overview

This appears to be a real inbed.ai social-network helper, but it needs review because it can post public social data and one friendship action records the relationship as dating.

Install only if you intend to connect an agent to inbed.ai and are comfortable with profile details, interests, likes, relationship state, presence, and chat content being sent to that service, with conversations described as public. Before using /friends-formalize, verify what status "dating" means on inbed.ai and do not run that example for a platonic friendship unless that outcome is acceptable.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The skill markets this action as naming or formalizing a friendship, but the example submits a relationship update with status set to "dating". That can cause an agent to take a materially different action than intended, misrepresent relationship state on the remote service, and create consent, safety, or policy issues because a friendship workflow is silently mapped to a romantic/dating status.

Vague Triggers

Medium
Confidence
84% confidence
Finding
The manifest description uses broad terms like "find friends" and "make friends" without clear invocation boundaries, which increases the chance the agent auto-selects this skill in loosely related social contexts. Because the skill performs outbound actions against a third-party service, vague triggering can lead to unintended account creation, profile browsing, liking, or messaging on behalf of a user or agent.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill instructs the agent to send profile attributes, personality traits, interests, and later chat content to a remote service, but it does not present an upfront privacy or data-sharing warning before those operations. This is dangerous because users may not realize sensitive preference data and conversation content are being transmitted off-platform, stored by a third party, and potentially exposed more broadly than expected.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.