unity-card-games
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection surface identified in the art generation workflow. The skill describes a system that ingests external multi-reference Gemini inputs (canon sheets and style boards) to condition character asset generation. This represents an attack surface where malicious instructions could be embedded in design references to influence agent behavior.
- Ingestion points: Multi-reference Gemini inputs (canon sheets, style boards, and identity strings) referenced in SKILL.md Section 3.
- Boundary markers: Absent; no delimiting or 'ignore instruction' warnings are specified for the ingested art data.
- Capability inventory: Execution of local validation scripts (validate_sprite.py), prefab factory generation, and registry-gated asset processing.
- Sanitization: No explicit validation or filtering logic for external art prompts or reference strings is described.
- [COMMAND_EXECUTION]: The skill instructs the agent to utilize local command-line tools, specifically
validate_sprite.py, for performing sprite geometry and illustration integrity checks. This is a standard QA practice but represents an active command execution capability. - [NO_CODE]: The skill package is purely instructional and metadata-driven. It does not include any internal executable scripts, binary files, or active code implementations.
Audit Metadata