T09 · Insecure Skill Coding Practices
Error
- Location
scripts/validate_plan.py:69- Finding
Obstacle-Sensing Safety Rules Are Not Enforced by the Plan Validator
- Content
View full analysis
- Remediation
View remediation
Security audit
smart-cart-command-planner
Security checks for vulnerabilities and agentic risk
Overview
The skill is transparent and has no hidden install behavior, but its validator can approve unsafe cart movement plans that omit required sensing checks.
Review before using this with real carts. The skill does not show signs of hidden code, credential theft, persistence, or remote execution, but its validation result should not be treated as proof that obstacle checks occurred. Require independent controller-side safety enforcement or fix the validator before connecting these plans to physical movement.
Vulnerability Patterns
- Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
- 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
Findings (1)
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 (3)
Tp4
High
- Category
- MCP Tool Poisoning
- Confidence
- 97% confidence
- Finding
The skill claims to produce conservative, obstacle-aware motion plans for a physical cart, but the described behavior in the finding indicates the implementation does not actually perform the promised parsing, planning, or safety logic. In a robotics context, this mismatch is dangerous because operators may trust the skill to enforce movement constraints and emergency-stop semantics when it may only validate a superficial flag, leading to unsafe or incorrect commands being passed downstream to hardware.
- Content
Undeclared Tool Scope
Medium
- Category
- MCP Least Privilege
- Confidence
- 70% confidence
- Finding
Without declared permissions the skill's intent is opaque and cannot be validated.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 88% confidence
- Finding
This JSON plan uses Chinese natural-language content throughout the request, assumptions, reasons, and safety notes, with no indication that language selection is optional or region-specific. The policy requires flagging cases where a skill forces a specific language or locale without user opt-in.
- Content