other
Warning
- Location
SKILL.md:129- Finding
Unsafe Self-Injury Technique Presented as an Emotional-Regulation Strategy
- Content
View full analysis
- Remediation
View remediation
Security audit
Rong Bing
Security checks for vulnerabilities and agentic risk
Overview
This markdown-only customer-service persona has no code or system access, but it needs review because it includes broad activation, crisis-adjacent emotional guidance, and one unsafe physical coping instruction.
Review before installing. Use this only as a fictional customer-service or roleplay training persona, preferably with explicit user opt-in. Do not treat it as therapy, crisis counseling, abuse-support guidance, legal advice, or a workplace mental-health protocol. The unsafe palm/fingernail coping instruction should be removed or replaced with non-injurious grounding before broad use.
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
Findings (1)
Vulnerability Patterns
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- 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 (7)
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The trigger phrases include broad everyday terms such as '客服', '安抚', and '投诉处理', which can cause the skill to activate unintentionally in normal conversations. In this skill, accidental activation is more concerning because it steers users into emotionally manipulative or quasi-counseling behavior without explicit consent, increasing the chance of inappropriate influence in sensitive situations.
- Content
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The skill is framed around emotional de-escalation and explicitly references domestic abuse hotline experience, trauma, and anger dissolution, yet provides no warning that it may engage in psychologically sensitive, crisis-adjacent interactions. This can mislead vulnerable users into treating the persona as qualified support, and the rich backstory plus authoritative tone make that risk higher in context.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The file content is entirely written in Chinese and the skill metadata also uses Chinese trigger phrases, which strongly constrains interaction to a specific language without any visible user opt-in or fallback. This can override user language preference, reduce transparency, and cause misunderstandings or inaccessible behavior for users who do not read Chinese.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 90% confidence
- Finding
This markdown file presents the skill content only in Chinese and does not indicate that language selection is optional or region-specific. The policy for natural-language violations requires flagging content that forces a specific language without user opt-in.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 84% confidence
- Finding
Mandating a specific language/locale without user choice can create safety and reliability issues by reducing user comprehension, increasing misunderstanding in already emotional conversations, and making consent to the role's methods less clear. In a de-escalation skill, forcing Chinese phrasing and tone conventions may also mis-handle users from other language contexts or cause the persona to continue despite a mismatch.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 90% confidence
- Finding
The trigger conditions are broad enough that this persona can activate for a wide range of emotionally charged interactions, including sensitive situations where the model may begin profiling, labeling emotions, or steering the user without explicit consent. In this skill, that risk is heightened because the role is designed to infer internal state automatically and to guide the conversation using influence-oriented tactics, which can cause inappropriate application outside narrow customer-service simulations.
- Content
Natural-Language Policy Violations
Low
- Category
- Not specified by scanner
- Confidence
- 78% confidence
- Finding
The file specifies a Chinese cultural and language context, including “当代中国二三线城市” and “母语普通话,” as part of the skill’s operating persona. The document does not indicate that users may choose another language or locale, which can conflict with language/locale choice expectations unless explicitly justified as a region-specific skill.
- Content