start
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process several repository-local files which could contain untrusted instructions if the repository is sourced from an external or untrusted party.
- Ingestion points: Reads
AGENTS.md,CLAUDE.md,Assets/CLAUDE.md,design/gdd/game-concept.md, and other project documents during the onboarding phase. - Boundary markers: None explicitly defined in the instructions to separate project data from system instructions.
- Capability inventory: The skill has access to tools like
Write(specifically forproduction/stage.txtandproduction/review-mode.txt) and broader file system interaction tools (Read,Glob,Grep). It also indicates it can perform direct code implementation for clear requests. - Sanitization: No explicit sanitization or instruction-ignoring logic is present for the markdown files it parses.
Audit Metadata