lovstudio-xbti-creator
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill clones the core project architecture from the developer's public repository (github.com/lovstudio/XBTI.git). As this is the official template source and hosted on a well-known service, this is a standard and expected operation.
- [COMMAND_EXECUTION]: Automates environment preparation by installing the lovstudio-image-creator skill via npx and configuring Python libraries (google-genai, Pillow). It uses the --break-system-packages flag for compatibility, which is a functional installation method for agent environments.
- [COMMAND_EXECUTION]: Utilizes node -e to programmatically merge generated Javascript data files into a single JSON manifest. This facilitates the transition from AI-generated content to a deployable static site structure.
- [PROMPT_INJECTION]: The skill exhibits a potential surface for indirect prompt injection where user-provided themes and tones influence the generation of Javascript files (dimensions.js, questions.js, types.js). 1. Ingestion points: User input for theme, BTI name, and style preferences captured via AskUserQuestion. 2. Boundary markers: None explicitly defined to separate user input from the code generation prompt. 3. Capability inventory: The generated files are later executed or required by node -e and require() calls during the packaging step. 4. Sanitization: Content is structured by the LLM into specific JS formats; folder names are sanitized using tr, but the generated JS content itself is not validated before execution.
Audit Metadata