ClawHub

Volcengine RDS PostgreSQL

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed Volcengine RDS PostgreSQL helper that uses the user's cloud keys for live read-oriented RDS and VPC queries.

Install only if you are comfortable letting the agent query Volcengine RDS and related VPC metadata. Use a dedicated least-privileged AK/SK, avoid broad production admin credentials, treat returned instance/account/configuration output as sensitive, and consider pinning the SDK version for repeatable use.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (2)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill explicitly requires sensitive environment variables (`VOLCENGINE_ACCESS_KEY` and `VOLCENGINE_SECRET_KEY`) and instructs the agent to invoke a local script that will consume them, but the skill does not declare explicit permissions governing access to those secrets. This creates a real secret-exposure and overreach risk: an agent using the skill may access cloud credentials without a clear permission boundary, and any future expansion of the script could use those credentials for unintended read or write operations against the user's cloud environment.

Description-Behavior Mismatch

Medium
Confidence
89% confidence
Finding
The skill is described as an RDS PostgreSQL operations assistant, but it also initializes and exposes VPC and subnet enumeration APIs. This expands the effective privilege and information-disclosure scope beyond the declared purpose, which can enable unintended cloud network reconnaissance and violate least-privilege expectations for users or orchestrators that trust the manifest.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal