improve-environment
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill performs legitimate environment auditing and improvement tasks within a Unity development context using standard tools.
- [PROMPT_INJECTION]: The skill ingests local knowledge files to build context for a sub-agent, creating an indirect prompt injection surface. * Ingestion points: Reads CLAUDE.md and architecture documentation in .ai/knowledge/. * Boundary markers: Content is interpolated directly into the sub-agent prompt without specific delimiters. * Capability inventory: The dispatched agent can read/write scene files and commit changes via git. * Sanitization: No sanitization is performed on ingested file content.
- [COMMAND_EXECUTION]: Uses git for version control and UnityMCP for scene interaction. These operations are restricted to the local project and require user confirmation.
- [DATA_EXFILTRATION]: Reads local project files to provide context. This data access is local and does not involve transmission to external domains.
Audit Metadata