Security audit
Security checks for vulnerabilities and agentic risk
This is a focused English-to-Chinese translation skill with disclosed quality-review and preference-memory behavior, and no evidence of hidden execution, exfiltration, or destructive actions.
Install this only if you want a Chinese-focused translation workflow. Before use, confirm whether the companion wenqu-review skill is installed and trusted, and treat writes to per-article preferences.md files as project changes that should be reviewed like any other repository edit.
The description repeatedly states that English materials should be translated into Chinese and lists activation on phrases like “translate this README” and “translate this paper,” including English requests that do not explicitly ask for Chinese. This imposes a language choice without offering the user a locale or output-language option.
The manifest says to use this skill when the user says phrases like “翻译一下” or “translate this paper,” which are broad requests that could match many contexts without clearly distinguishing when this specific skill should or should not activate. Although some examples are given, the trigger scope is still open-ended and lacks negative examples or tighter constraints.
The skill directs cross-skill invocation and, failing that, direct reads from another skill's local reference path in the user's home directory. That expands the skill's access beyond its stated translation purpose, breaks isolation assumptions between skills, and could expose private prompt assets or internal guidance not intended for this skill's trust boundary.
The skill explicitly instructs persistent writes to per-article preference files under the project tree, which exceeds a translation-only function and creates stateful side effects from user content. This can leak or retain sensitive preferences/content across sessions, modify user repositories unexpectedly, and enable prompt-injection persistence if hostile source text or feedback is recorded as a durable preference.
The manifest description states the skill translates materials '成中文', and the document consistently instructs producing Chinese output as the required behavior. This is a natural-language locale policy concern because it mandates a specific language rather than offering a user choice or documenting a justified region-specific constraint.
This line instructs the agent to rewrite content into '自然中文' as a fixed requirement. Because the file does not offer an explicit language/locale choice at this point, it reinforces a forced-language behavior covered by the policy-violation category.
The instruction says the content should read as if written by a native Chinese writer, which is appropriate only after the user has selected Chinese. As written, it contributes to a mandatory language constraint without opt-in or documented justification.
No suspicious patterns detected.
