Security audit
Security checks for vulnerabilities and agentic risk
This is a disclosed Markdown-only advisory skill for applying a Paul Graham-style perspective, with no code execution, persistence, credentials, or data access.
Installers should know this skill is designed around a Chinese-language PG advisory style. Ask for another language explicitly if needed, and treat its Paul Graham lens as opinionated framing rather than current factual authority.
The skill instructs the agent to default to Chinese output ('默认用中文输出') without checking the user's language preference. This can override user expectations or system defaults, causing instruction-priority conflicts and reducing usability or clarity, especially in multilingual or safety-sensitive contexts where precise comprehension matters.
The default contract explicitly mandates Chinese as the output language ('输出语言：中文') without offering user choice. In an agent setting, hard-coded language defaults can interfere with higher-priority user intent, create avoidable confusion, and in some workflows may cause downstream misinterpretation if operators or tools expect another language.
This markdown file presents all substantive guidance in Chinese, with no indication that the user can choose another language or that the skill is intentionally limited to a Chinese-language audience. Under the policy, forcing a specific language without opt-in is a natural-language policy violation.
The file explicitly defines the skill’s expression style entirely in Chinese, which can steer model output into a specific language without any user opt-in or documented locale requirement. This is not a classic security exploit, but it is a genuine policy/control issue because it can override user expectations, reduce accessibility, and create prompt-injection-like behavior at the style layer.
This manifest presents the skill as Chinese-language-facing ('Paul Graham 视角' and Chinese short description) while the default prompt hard-codes English instructions. That can impose a language behavior without explicit user choice, which fits the locale/language policy concern for natural-language policy violations.
This markdown file contains substantive instructional content in Chinese under English headings, but does not indicate that the skill is intentionally Chinese-language or that users can choose their preferred language. That can violate language/locale policy when a skill presents guidance in a fixed language without opt-in or justification.
No suspicious patterns detected.
