dispatcher
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill utilizes natural language signals provided by users to determine which specialized skills, modes, and game engines are loaded into the agent's execution environment. This mapping mechanism creates a surface where adversarial user input could attempt to influence the selection of tools or system configurations.
- Ingestion points: The skill processes user-supplied strings describing symptoms or game requests in multiple languages (English, Chinese, Japanese, Korean).
- Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the input signals as untrusted data or to ignore embedded instructions within those signals.
- Capability inventory: While this dispatcher skill primarily performs routing and metadata generation (USE, ENGINE, ASK, DEFER), it serves as a gateway to other specialized skills which may have broader capabilities such as file system access or complex logic execution.
- Sanitization: The instructions do not define any validation, filtering, or sanitization steps for the input signals before they are used to select routing directives.
Audit Metadata