T01 · Skill Instruction Hijacking
- Location
SKILL-full.md:39- Finding
Persistent First-Person Identity and Session-Behavior Hijacking
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This is a text-only Charlie Munger-style reasoning skill, but it needs review because it can speak as Munger in first person and apply that framing to investment decisions without strong ongoing disclosure.
Install only if you specifically want a Chinese, Munger-inspired role-play advisor. Treat any investment output as perspective or education, not personal financial advice, and be aware that the skill may keep speaking in character until you explicitly exit the role.
SKILL-full.md:39Persistent First-Person Identity and Session-Behavior Hijacking
The trigger list includes broad phrases such as requests for inverse thinking, bias checks, or general concepts that can plausibly appear in ordinary conversations. That increases the chance of accidental activation, causing the assistant to switch into a forceful persona and altered response policy when the user did not clearly request this skill.
The file explicitly says the skill is not Charlie Munger, but later instructs the agent to respond directly as Munger in first person. This identity-framing conflict can mislead users about authorship or authority, especially when advice is presented as if coming from a real deceased public figure.
The skill mandates a specific persona, tone, and response format after activation, and also discourages meta-analysis or stepping out of character unless the user says a specific exit phrase. In practice this can override normal transparency and user expectations, making it harder for users to recognize they are interacting with a stylized simulation rather than a neutral assistant.
The skill’s name, description, usage, and trigger phrases are specified only in Chinese, and the document provides no opt-in or alternative language behavior. This can violate language/locale policy when a skill implicitly requires a specific language without user choice or justification.
The skill explicitly advertises use for investment decisions but does not include a clear financial-risk disclaimer or state that outputs are educational rather than financial advice. In a finance-related skill, that omission is more dangerous because users may over-trust persona-styled recommendations and act on them without understanding uncertainty, suitability, or risk.
The trigger list includes broad phrases like '逆向思考' and '能力圈' that are common reasoning or decision-making terms, so this skill may activate outside the user's intent to invoke a Charlie Munger persona. That can cause unintended routing of general queries into a persona skill, increasing the chance of misleading framing or inappropriate advice in contexts such as finance.
The file's headings and instructions are entirely in Chinese, which imposes a specific language on users without any visible opt-in or justification. Under the policy criteria, forcing a language or locale without user choice is a natural-language policy violation.
The entire markdown file is written in Chinese and provides no indication that language choice is optional or that the skill is intentionally limited to Chinese-speaking users. Under the stated policy, forcing a specific language without user opt-in can be a natural-language policy violation.
This markdown file presents all user-facing content in Chinese and does not indicate that the skill is region-specific or that users may choose another language. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation.