Security audit
Security checks for vulnerabilities and agentic risk
This skill is a coherent template-based report generator, with no evidence of hidden execution, persistence, credential access, or unrelated behavior.
Before installing, treat any template, report, CSV, SQL result, or knowledge-base excerpt you provide as content that may be processed by the configured cloud model. Use only files you intend the skill to read, avoid highly sensitive data unless approved for cloud processing, and review generated documents before relying on them.
The instruction is written entirely in Chinese and directly addresses the model as a Chinese-language document generation assistant, which effectively forces a specific language/locale for outputs. Under the policy, language constraints are only acceptable when users are given an explicit choice or when the restriction is clearly documented and justified as region-specific.
The skill explicitly states that local files and locally stored data are processed while a cloud-hosted model ('glm-5') is used, but it does not disclose that user content may be transmitted off-host to the cloud service. This creates a real data exposure risk because users may supply sensitive templates, reports, CSVs, or SQL-derived data under the assumption that processing remains local except for tooling.
The skill description is overly broad: it claims to process Word templates, knowledge-base documents, multiple documents, CSV/relational tables/SQL query results, and automatically generate formal documents, but it does not define activation boundaries, permitted data sources, or user-consent constraints. In combination with powerful tools like read, write, and exec, this ambiguity can cause the agent to act on unintended inputs or sensitive data, expanding attack surface and increasing the risk of misuse or overreach.
The description is written as a Chinese-only capability statement and implies the skill operates in Chinese without offering a language or locale choice. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale restriction is clearly documented and justified.
The natural-language instructions and usage description are presented only in Chinese, which can amount to forcing a specific language without user opt-in. The file does not indicate that the skill is region-specific or provide any alternative language option.
No suspicious patterns detected.
