Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed book-recommendation helper that searches public book sites and runs a small local scoring script, with no evidence of hidden, destructive, or persistent behavior.
Install only if you are comfortable with your book topics and candidate titles being searched on third-party services. The skill is strongly Chinese/Douban-oriented and includes a local Python scoring utility, but the inspected utility is narrow and deterministic.
The documented search strategy hard-codes Chinese queries such as "经典书籍推荐" and "必读书单", and the example output is also presented in Chinese fields/content. This indicates a locale/language preference is imposed by default rather than offered as a user-selectable option, which is a natural-language policy concern under the language/locale rule.
The input examples, output schema values, search templates, and required warning/error text are predominantly fixed in Chinese, indicating a Chinese locale requirement. The file does not offer users a language/locale choice or explain that the skill is intentionally limited to a Chinese-language context.
The skill explicitly directs the agent to invoke a local script (scripts/score_books.py) and pass collected data through stdin, expanding the behavior from web-based recommendation into local code execution. Even if the script is intended to be deterministic, introducing executable local components increases attack surface because the script may be modified, unsafe, or process untrusted input in unexpected ways.
This code file contains natural-language comments and later user-visible error strings in Chinese, while the rest of the file is in English. Under the policy, forcing a specific language without opt-in is a locale/language policy violation unless the constraint is documented and justified.
The exception messages returned to users are only in Chinese, with no mechanism for language selection or indication that the skill is region-specific. This is a direct natural-language policy issue because it constrains the user experience to one language by default.
The skill instructs sending user-supplied topics and candidate book queries to external search engines and sites such as Douban and Goodreads without disclosing that this data leaves the local environment. This creates a privacy and transparency issue, especially if topics reveal sensitive interests, internal research areas, or user behavior patterns.
No suspicious patterns detected.
