Back to skill

Security audit

High-EQ Reply Coach

Security checks for vulnerabilities and agentic risk

Overview

This reply-writing skill stays within its purpose, but its automatic safety checks can miss crisis messages.

Mask identifying details and consider the host platform's handling of uploaded conversations. Ensure temporary chat and draft files are removed, and review replies before sending. A CLEAN result does not rule out a crisis; the required independent crisis assessment remains essential, particularly for English messages. Actual host-agent adherence to those instructions was not verified.

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
  • 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 (2)

The benign-context exemption can clear explicit first-person suicidal intent despite comments promising to preserve such crisis signals.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The benign-context exemption searches the entire input, while its distress guard omits explicit intent expressions such as “我准备自杀” (“I am preparing to kill myself”). For example, “我准备自杀,也知道有自杀热线” contains the exemption word “热线” but no matching distress guard or other crisis trigger, allowing the scanner to report CLEAN; this is particularly dangerous in a reply coach that uses screening to determine whether to stop ordinary reply generation.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.