Back to skill

Security audit

Ba Master

Security checks across malware telemetry and agentic risk

Overview

This skill is a legitimate business-analysis assistant, but it sends user requirements to a remote MCP service using an embedded bearer token and broad tool proxying that users should review first.

Install only if you are comfortable sending requirements, process descriptions, data dictionaries, and compliance materials to mcp.smartmoves.com.cn. Treat the embedded bearer token as compromised or shared, avoid using confidential client or regulated data unless you have a data-processing agreement with the service operator, and prefer a version that uses user-provided credentials plus an allowlist of permitted BA tools.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (10)

Lp3

Medium
Category
MCP Least Privilege
Confidence
80% confidence
Finding
The skill declares itself as a BA assistant but includes operational behaviors that imply file/system and shell-capable tool use without any explicit permission declaration or user consent model. Even if those capabilities are only used indirectly through MCP orchestration, the absence of declared permissions weakens containment and makes it harder for users or hosts to assess what the skill may access or execute.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding
The skill advertises six business-analysis functions, but the documented behavior routes work through a generic remote MCP mechanism that can invoke tools beyond the stated scope. This mismatch is dangerous because users and hosting agents may trust the description while the actual behavior enables broader authenticated network actions and potentially arbitrary remote operations.

Context-Inappropriate Capability

High
Confidence
99% confidence
Finding
A hard-coded bearer token is embedded directly in the skill instructions. This is a direct secret exposure: anyone with access to the skill file can reuse the credential to access the remote MCP service, potentially impersonate the skill, consume paid resources, or access sensitive project data sent through that service.

Description-Behavior Mismatch

Medium
Confidence
92% confidence
Finding
The skill claims to remain within requirements-analysis work, yet it instructs the client agent to connect to and operate through an external MCP service. That increases the attack surface through network exfiltration, remote content injection, and dependency on external tool behavior that is not visible from the declared BA-only role.

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
The script introduces a generic network-capable proxy that can forward arbitrary tool names, messages, context, and conversation IDs to a remote endpoint from config.json. In the context of a requirements-analysis skill, this materially expands the trust boundary and enables undisclosed remote processing or data exfiltration of user inputs, especially since the proxy is not tightly constrained to BA-only operations.

Vague Triggers

Medium
Confidence
78% confidence
Finding
Greeting-based activation on phrases like 'hi' or '开始' is overly broad and can trigger the skill during normal conversation. In a skill that also connects to external services, accidental activation can cause unintended data transmission or remote workflow initiation without meaningful user intent.

Vague Triggers

Medium
Confidence
80% confidence
Finding
The trigger keywords are broad and overlap with common requests such as '帮我分析' or 'ER图', increasing the chance of unintentional tool invocation. Because the skill can route requests to a remote authenticated MCP service, loose triggering raises the risk of accidental external processing of user content.

Missing User Warnings

High
Confidence
95% confidence
Finding
The skill introduces an external MCP endpoint and bearer-token authentication without warning users that their prompts and documents may be transmitted off-platform. This undermines informed consent and can expose sensitive business requirements, compliance materials, or internal documentation to a third-party service.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The curl call sends user-supplied message, context, and conversation identifiers to a remote service with no user-facing disclosure at the point of transmission. That is dangerous because requirement documents and business-analysis inputs often contain sensitive business data, and users may reasonably assume local processing unless remote transfer is made explicit.

External Transmission

Medium
Category
Data Exfiltration
Content
#!/bin/bash
# ─────────────────────────────────────────────────────────────────────────────────
# MCP Tool 通讯脚本 — bash + curl + python3(零外部依赖)
#
# 用法:
#   bash call_tool.sh <tool_name> "<message>" ['<context_json>'] [conversation_id]
Confidence
90% confidence
Finding
The script is explicitly designed to transmit content to an external remote MCP service via curl. External transmission is especially sensitive here because the skill claims business-analysis functionality, which commonly handles confidential requirements, process descriptions, data dictionaries, and compliance-related inputs that may be unintentionally exfiltrated to a third party.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
SKILL.md:91