Tp2
High
- Category
- MCP Tool Poisoning
- Confidence
- 85% confidence
- Finding
Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.
- Content
Security audit
Global Customer Agent
Security checks for vulnerabilities and agentic risk
Overview
This documentation-only skill is coherent, but it should be reviewed because it encourages automatic customer replies and price quotes without clear approval or data-handling limits.
Review this before connecting it to live customer channels, pricing, inventory, or CRM systems. Require human approval for quotes and sensitive replies, define what customer data it may access, and add fallback rules for uncertain language detection, pricing errors, and mistranslations.
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
Vulnerability Patterns
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
- MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
- Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 86% confidence
- Finding
The skill explicitly promises automatic multilingual customer reception, product recommendation, inventory querying, and quote generation, but provides no safeguards around customer-data handling, review requirements, or controls for outbound business communications. In a customer-service and quoting context, this can lead to unauthorized disclosure of customer or pricing data, inaccurate quotes, and uncontrolled automated messages affecting privacy, contracts, or commercial commitments.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 74% confidence
- Finding
The skill describes automatic handling across 36 languages without stating that users can choose a language, opt out, or confirm detected language before business communication is sent. While not inherently malicious, forced auto-detection in a sales/support setting can cause miscommunication, unintended disclosure, or sending incorrect commercial content to customers in the wrong language.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 78% confidence
- Finding
The manifest presents the skill name, description, and tags entirely in Chinese while advertising support for 36 languages. This creates a natural-language locale constraint without stating that users may choose their preferred language or that the Chinese-only metadata is intentional for a region-specific audience.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 86% confidence
- Finding
The '语言优先级' section assigns fixed default support levels such as 'P0' for English and Chinese and lower priorities for other languages. This can be interpreted as a language-policy preference without stating that users may choose their preferred language, which risks violating locale-choice expectations.
- Content
Static analysis
No suspicious patterns detected.