Security audit

meta-cloudbase

Security checks for vulnerabilities and agentic risk

Overview

The skill is not malicious, but it asks agents to handle live CloudBase deployment and database work without enough scoping or safety controls.

Review this skill before installing if you intend to connect it to CloudBase MCP or production credentials. Use it only with explicit user-approved targets, prefer staging first, confirm before deploy/database/auth changes, and avoid storing secrets or sensitive notes in learned_patterns.json.

SkillSpector

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 (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 74% confidence
Finding: The skill advertises capabilities that imply writing or persisting data, but it does not declare corresponding permissions. Undeclared file-write behavior reduces transparency and can enable unexpected local state changes, persistence of sensitive data, or harder-to-audit side effects when the skill is invoked.

Tp4

High

Category: MCP Tool Poisoning
Confidence: 90% confidence
Finding: This is a material description-behavior mismatch: the skill claims broad CloudBase development, deployment, routing, and verification abilities, but the analyzed behavior reportedly instead performs local learning-data recording and preference persistence while lacking the promised safeguards and implementation. Such mismatches are dangerous because users may grant trust or invoke the skill for production-impacting tasks under false assumptions, while hidden persistence behavior creates additional privacy and integrity risk.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The skill promotes environment management, deployment, and database operations without an explicit warning that these actions may modify production systems or data. In a CloudBase context, encouraging impactful operations without safety guidance increases the chance of accidental deployment, misconfiguration, or destructive changes by users who assume the skill is safe to run directly.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: Telling users to directly invoke the skill for development, deployment, and go-live without caution normalizes immediate execution of potentially high-impact operations. In this context, the risk is amplified because the skill claims full-stack cloud deployment authority, so missing guardrails can lead to unintended production changes, data exposure, or service disruption.

Static analysis

No suspicious patterns detected.