team-level
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill reads from user-controlled design documentation files and passes that context to subagents with write and bash capabilities.
- Ingestion points: Project files are read from
design/gdd/,design/levels/, anddesign/narrative/. - Boundary markers: No explicit delimiters are used in the prompts sent to subagents via the Task tool.
- Capability inventory: Subagents are granted access to powerful tools including
Bash,Write, andEdit. - Sanitization: External content is not sanitized before being passed to subagents.
- Mitigation: The workflow is governed by strict human-in-the-loop (HITL) gates at every step using
AskUserQuestion, ensuring the user reviews subagent proposals before execution. - [SAFE]: The orchestration logic is transparent, and all potential risks are intrinsic to the game design task and sufficiently mitigated by manual oversight.
Audit Metadata