ClawHub
Back to skill

Security audit

mayu-en

Security checks for vulnerabilities and agentic risk

Overview

This is an instruction-only reflection skill that can broadly shape the agent's behavior, but its behavior is visible, coherent with its purpose, and not backed by executable code or hidden data access.

Install this only if you want a global reflection framework to influence the agent's normal behavior. If your agent has memory or file-editing tools, keep those scoped to approved memory locations and require clear user intent before reading, changing, or deleting stored memories.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Confidence
92% confidence
Finding
The trigger keywords are broad, generic phrases like "AI behavior" and "self-reflection" that can easily appear in ordinary conversation, increasing the chance this skill activates unintentionally. Because the skill is also configured as a high-weight global behavioral layer, accidental activation could alter responses outside the user's intent and interfere with higher-priority safety or policy behavior.

Vague Triggers

Medium
Confidence
96% confidence
Finding
The skill is marked global, high-priority, and always-on, with no clear boundaries limiting when its behavioral instructions should apply. In practice, this means the prompt can continuously shape model behavior across unrelated tasks, creating persistent prompt-surface exposure and increasing the risk of policy drift, unintended instruction precedence, or interference with safer system behaviors.

Static analysis

Detected: suspicious.prompt_injection_instructions

Prompt-injection style instruction pattern detected.

Warn
Code
suspicious.prompt_injection_instructions
Location
SKILL.md:209