higgsfield-game-generation
Fail
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill downloads an installation script from 'https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh' and pipes it directly to the shell. This is intended to install the Higgsfield CLI, the core tool the skill utilizes. While this is a common installation pattern for this vendor, executing remote scripts always carries inherent risks.
- [EXTERNAL_DOWNLOADS]: The skill downloads Blender from 'blender.org' and retrieves 3D assets from public GitHub repositories. These downloads are from reputable sources and are required for the 3D asset processing pipeline.
- [COMMAND_EXECUTION]: The skill performs extensive shell command execution using the Higgsfield CLI, 'blender', 'ffmpeg', and several bundled Python scripts to automate game development tasks.
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by using untrusted user input to construct prompts for downstream AI models (image, video, and 3D generation). \n- Ingestion points: User game briefs and asset requests provided via 'SKILL.md' arguments. \n- Boundary markers: None; user descriptions are interpolated directly into prompts for models like 'flux_2' and 'seedance1_5'. \n- Capability inventory: Shell execution of the 'higgsfield' CLI and various local scripts for model manipulation. \n- Sanitization: No input sanitization or validation is performed on the user-supplied strings before they are used in generation prompts.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh - DO NOT USE without thorough review
Audit Metadata