ClawHub
Back to skill

Security audit

aws

Security checks across malware telemetry and agentic risk

Overview

This AWS skill is broadly coherent, but it under-labels live AWS-changing commands as read-only and uses overly broad activation keywords.

Install only if you want an agent to help with AWS infrastructure work. Treat every AWS or Terraform command as potentially able to change real cloud resources or incur cost, and require the agent to explain the target account, region, resource, and whether the command is read-only before execution.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

Medium
Confidence
97% confidence
Finding
The document states that all shown CLI commands are read-only by default, but it also includes mutating commands such as `aws ec2 create-vpc-endpoint` and `aws logs put-retention-policy`. This mismatch can mislead an agent or user into executing state-changing operations under a false safety assumption, which is especially risky in an AWS skill where commands affect live cloud resources and cost/security posture.

Vague Triggers

High
Confidence
89% confidence
Finding
The trigger keywords are very broad (`web`, `deploy`, `architect`, `optimize`, `infrastructure`, `amazon`) and can match many unrelated user requests. That increases the chance this skill activates out of scope and steers an agent into AWS-specific guidance or command suggestions in contexts where cloud changes were not intended.

Vague Triggers

High
Confidence
88% confidence
Finding
The repeated trigger keyword section remains vague and broad, reinforcing the same over-activation problem elsewhere in the file. Duplicating broad triggers can make accidental invocation more likely and increases the chance that AWS operational advice is injected into unrelated conversations.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.