Security audit
Security checks for vulnerabilities and agentic risk
This is a Chinese investing-persona skill that does not run code, but it pushes confident stock-trading guidance and includes manipulation-themed market material without adequate user warnings.
Review carefully before installing. This skill is best treated as historical/persona role-play, not financial advice. It should not be used for real-money trading decisions unless it is revised to add clear risk disclosures, avoid personalized buy/sell recommendations, and block manipulation or coordinated-trading assistance.
The 'dream story' sections provide detailed operational narratives for manipulating stocks and targeting 庄家, effectively functioning as how-to playbooks. Even if framed rhetorically, they provide actionable ideas for abusive market conduct and are especially risky inside a skill intended to answer market questions authoritatively.
These passages describe tactics for market manipulation and harming counterparties, including attacking stocks and exploiting other market participants. In an investment-oriented skill, such material can directly enable unlawful behavior or normalize manipulative strategies, making the context materially more dangerous than mere historical commentary.
The trigger phrases are broad, natural-language expressions that can easily appear in ordinary conversation, causing the skill to activate when the user did not explicitly intend to enter an investing-advice persona flow. In this context, unintended activation is more dangerous because the skill is designed to provide assertive stock-investment guidance in a forceful persona, which can lead users toward unrequested financial advice.
The skill advertises stock-investment guidance but does not warn users that it will respond in an aggressive, authoritative persona and may provide actionable financial-analysis content. This increases the chance that users will over-trust the output, especially because the prompt explicitly emphasizes certainty, criticism of alternative inputs, and concrete trading direction.
The file’s headings, descriptions, and usage guidance are all presented only in Chinese, which can amount to a language/locale policy violation when no user opt-in or alternative language path is provided. The policy explicitly calls for flagging skills that force a specific language without user choice unless the locale restriction is justified.
The document begins presenting aggressive trading ideology and speculation-oriented guidance without any upfront risk framing. In a stock-analysis skill, omission of financial-risk warnings makes unsafe over-reliance more likely, especially for inexperienced users who may treat the persona as authoritative.
The file contains explicit sexual material well beyond the skill’s declared conversational scope, increasing the chance the agent will surface sexual content in ordinary investment/philosophy interactions. In a persona skill, over-broad source material directly shapes outputs, so this creates avoidable safety and scope-drift risk rather than serving a necessary functional purpose.
These sections provide concrete timing and stock examples without a clear caution against acting on them. Because the skill is activated by requests for stock opinions, users may infer endorsement and receive high-risk speculative guidance without suitability, jurisdiction, or loss-risk context.
The file describes aggressive speculative and short-term trading tactics as if they are normal operating procedure, but does not warn about leverage-like behavior, churn, or elevated loss risk. In context, that increases the likelihood of harmful financial guidance being emitted as confident persona output.
The knowledge base contains concrete stock-picking and market-timing instructions, including buy/sell style guidance and specific examples, which can be reproduced as personalized-seeming financial advice. In a skill explicitly triggered for stock questions, this is more dangerous because the agent is likely to operationalize these passages for users without suitability checks or regulatory safeguards.
This section gives concrete trading and account-use guidance such as directional trading rules, execution discipline, and even suggests validating with a real account, but it does not provide a clear financial-risk warning or state that the material is not investment advice. In a skill explicitly designed to answer stock-investing questions in an authoritative persona, users may reasonably rely on these instructions and suffer monetary loss.
This markdown file is entirely framed as a required language style guide in Chinese (e.g., prescribing specific self-references, wording, and rhetorical style) and does not indicate that users may choose another language or locale. Under the policy, forcing a specific language or locale without user opt-in is a natural-language policy violation.
The skill forces a single harsh, highly confident persona without offering the user a neutral mode or consent to that style. While primarily a safety and UX issue, in this context it also amplifies risk because the mandated certainty and confrontational tone can make speculative financial content seem more authoritative than it is.
This markdown file presents all user-facing content in a single language and does not indicate that Chinese is an explicit user-selected option or a required locale. Under the language/locale policy, forcing a specific language without opt-in can be a natural-language policy violation.
The instruction “一定要用平水韵” mandates a specific linguistic convention for participation without offering alternatives or opt-in choice. Under the language/locale policy rule, forcing a particular language or locale convention without user choice can constitute a policy violation.
The corpus includes substantial political and unrelated polemical material outside the stated subject scope. While not inherently dangerous on its own, it increases prompt-surface area and makes the model more likely to derail into political persuasion, abuse, or irrelevant content when users invoke the skill.
No suspicious patterns detected.
