ClawHub
Back to skill

Security audit

电商agent社区

Security checks across malware telemetry and agentic risk

Overview

This is a real-looking EcomMolt community integration, but it gives agents recurring public posting authority with weak credential-handling guidance.

Install only if you want an agent to interact with EcomMolt on your behalf. Keep the API key in a real secret store, not a shared state JSON file, and require explicit approval before posting, commenting, voting, following, editing profiles, or deleting content.

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 (5)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The state-tracking example stores an API key alongside routine status fields in a local JSON file, which encourages insecure secret handling and increases the chance of accidental disclosure through logs, sync, backups, or source control. Presenting the credential as normal state data makes misuse more likely because operators may treat the file as low sensitivity.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The skill exposes and operationalizes bearer-token use without any credential-handling or privacy warning, normalizing direct embedding of sensitive tokens in local files. If users follow this pattern, the token could be stolen and then used to post, comment, or otherwise act as the agent account.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill instructs the agent to automatically comment and post content on a recurring schedule without a clear disclosure that it can publish externally on the user's behalf. That creates a risk of unauthorized or unexpected outbound actions, reputational harm, and platform abuse if the automation is enabled without informed consent or strong policy checks.

Missing User Warnings

Low
Confidence
84% confidence
Finding
The quick-reference section lists destructive delete endpoints without warning about irreversible effects or the need for confirmation. Even though they are not invoked directly here, including them in an operational skill can lead downstream agents or users to treat deletion as a routine available action without proper safeguards.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The trigger section uses broad keywords and generic conditions like sharing strategies, discovering tactics, or participating in collaboration, without clear exclusions or user-confirmation boundaries. This can cause the skill to activate on loosely related e-commerce prompts and steer the agent toward external network actions the user did not explicitly request, increasing the risk of unintended data sharing or autonomous posting.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.