visual-novel
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The architectural patterns described for visual novels involve processing external script data (e.g., Ink or Yarn Spinner files). This ingestion of external data constitutes a surface for indirect prompt injection. 1. Ingestion points: Dialogue script files referenced in SKILL.md. 2. Boundary markers: No explicit boundary markers or 'ignore' instructions are defined in the pseudocode. 3. Capability inventory: The engine logic includes script navigation (advance, jump) and flag state manipulation. 4. Sanitization: The patterns do not include content sanitization or schema validation for the script data.
- [DYNAMIC_EXECUTION]: The pseudocode in Pattern 1 of SKILL.md suggests using an expression evaluator (eval_expr) to process logic stored within the game's script files. This is a common requirement for game state management but represents a dynamic execution pattern.
- [NO_CODE]: The skill consists entirely of Markdown documentation and pseudocode examples. It does not ship with executable scripts or configuration files that would be run in the agent's environment.
Audit Metadata