ClawHub
Back to skill

Security audit

Fanbozhe Clean

Security checks for vulnerabilities and agentic risk

Overview

This is a conversational decision-review skill that gives blunt adversarial feedback, with no code execution, data access, persistence, or hidden install behavior found.

Install only if you want direct, sometimes harsh critique of ideas or plans. Avoid using it for crisis, mental-health, or emotionally vulnerable situations, and invoke the adversarial mode explicitly when you want that tone.

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
  • 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 (3)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill is explicitly designed to give '无麻醉结论' and to '骂醒我', but the description does not clearly warn users that outputs may be blunt, emotionally intense, or unsuitable for vulnerable users. In context, this is more dangerous because the skill's core behavior intentionally increases confrontation, which can lead to distress, misaligned expectations, or accidental harm if triggered implicitly.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list includes broad natural-language phrases such as '红队', '第二意见', and '反驳我' in metadata, which can plausibly appear in ordinary user requests without the user intending to invoke a harsh adversarial persona. That creates a prompt-routing risk: the skill may activate unexpectedly and produce more aggressive or oppositional behavior than the user anticipated.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file defines two modes but the switching condition is only described as '点名时' and later lists many overlapping trigger phrases, without a strict disambiguation rule. This ambiguity increases the chance that the assistant enters the adversarial mode based on loose semantic matching, causing tone escalation or unintended emotionally intense output.

Content

No source excerpt is available for this finding.