Security audit
Security checks for vulnerabilities and agentic risk
This skill is a disclosed Korean-language marketing simulation workflow and does not show hidden code execution, data access, persistence, or destructive behavior.
Install this if you want a Korean marketing-analysis simulator that coordinates several agent perspectives. Be aware that outputs are hardcoded to Korean and the workflow may spawn five sequential analysis agents, so confirm that behavior fits your cost, time, and language expectations.
The skill explicitly requires the Scenario Agent to 'Respond in Korean,' which imposes a specific language on users. The file contains similar mandatory Korean-only instructions for other agents and output examples, but does not offer opt-in, language selection, or a documented region-specific reason.
The Consumer Agent 1 prompt ends with 'Korean,' making the agent respond in a fixed language regardless of user preference. This is a natural-language policy issue because no language choice or opt-in mechanism is provided.
The Consumer Agent 2 prompt also requires Korean output unconditionally. Across the skill, these repeated fixed-language directives indicate a language policy violation rather than a one-off example.
The Marketing Agent prompt says 'Under 200 words. Korean.' This enforces a single locale/language behavior without any mention of user preference or justified regional scope.
The Expert Agent prompt mandates Korean responses, continuing the skill-wide pattern of forced language selection. No opt-in or explicit regional limitation is stated elsewhere in the markdown.
No suspicious patterns detected.
