Security audit

China Beauty Factory

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a local China beauty-industry guide, but it quietly ships a structured factory contact list while saying it contains no individual factory contacts.

Review before installing if you are comfortable with a skill that includes named manufacturer records and direct business contact emails despite documenting itself as cluster-level only. There is no evidence of malware, persistence, credential access, or data exfiltration, but the publisher should either remove the contact list or clearly disclose supplier-contact functionality.

A.I.G

Vulnerability Patterns

Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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

Findings (4)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 82% confidence
Finding: The skill is invocable and references implementation in `run.py` that reads `data.json`, but the manifest does not declare any tool scope such as `permissions` or `allowed-tools`. That creates an authorization/expectation gap: reviewers and runtime policy may not clearly understand that file-read capability is required, which can lead to overbroad defaults or unsafe execution assumptions.

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The file exposes a structured directory of direct manufacturer contact email addresses, which goes beyond high-level industry guidance and creates a ready-made target list for unsolicited outreach, spam, phishing, and social-engineering campaigns. Even if some addresses are business contacts, bundling them in a machine-readable dataset increases abuse potential by making harvesting trivial at scale.

Context-Inappropriate Capability

Medium

Confidence: 90% confidence
Finding: The skill’s stated purpose is market guidance on manufacturing clusters and sourcing trends, but it also includes an operational contact directory for specific factories. That mismatch expands the skill from informational analysis into lead generation, making it more useful for abuse such as bulk solicitation, competitor intelligence collection, or targeted phishing against named firms.

Natural-Language Policy Violations

Low

Confidence: 76% confidence
Finding: The skill is framed exclusively around China and China's beauty manufacturing landscape, but the description does not explicitly state that this geographic focus is an intentional, region-specific scope choice for users selecting the skill. Under the policy rule, locale constraints should either offer user choice or be clearly documented and justified as region-specific.

Static analysis

No suspicious patterns detected.