Security audit
Security checks for vulnerabilities and agentic risk
This is a simple local AI-vendor due-diligence checklist with documentation and localization issues, but no hidden network, credential, persistence, or destructive behavior.
Before installing, expect a local checklist tool with Chinese-first CLI output. Do not rely on the documented JSON-input path unless it is fixed; use the shown CLI arguments instead. Consider reducing unused declared tool access such as WebFetch if your environment enforces tool permissions tightly.
The file’s operational instructions and agent guidance are written entirely in Chinese, beginning with the title and continuing through all behavioral rules. Under the stated policy, forcing a specific language without user opt-in is a natural-language policy violation unless a locale constraint is clearly documented and justified, which is not present here.
The file prominently presents core instructions and policy text in Chinese, including licensing and disclosure content, without indicating that users may choose another language. This creates a natural-language policy concern because it imposes a specific language/locale experience without opt-in or a documented region-specific justification.
The main description field is written in Chinese, and several key metadata fields and operating instructions are also Chinese-first. This can impose a specific language on users without explicit opt-in or a documented locale restriction, which matches the language/locale policy concern.
The documentation says 'Vendor data can be passed via JSON input,' yet the manifest interface is strictly CLI args and the input schema lists only mode, vendor, score, threshold, out, and json-output flag fields. This is an active documentation-to-interface contradiction about how the skill accepts data, not just an omitted detail.
Line L107 says vendor-specific information may be passed 'through JSON input' for the script to read, while the manifest's interface is args-based and the schema contains no corresponding JSON input option. This contradicts the documented behavior of the skill's input handling.
The module docstring, usage text, CLI descriptions, and user-facing output all assume Chinese-only operation. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale constraint is clearly documented and justified, which it is not here.
The activation/use-context list is only provided in Chinese, which may constrain discoverability and usage to a specific language audience without saying this is intentional or optional. Because the skill is otherwise bilingual, this inconsistency can still amount to an implicit language preference.
