higgsfield-game-generation

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The list is mostly official downloads, API endpoints, and GitHub repos (low risk), but the raw.githubusercontent.com URL points to a shell install script that the skill pipes to sh — executing an unreviewed remote script is a high-risk pattern that can distribute malware.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The runtime workflow ingests outsider-authored free text via the user-facing “[game brief or asset request]” and other user-supplied fields during the “Plan” phase (e.g., STYLE FORMULA approval and asset prompts), which the agent then uses to construct generation prompts.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs runtime execution of remote installers/binaries (curl | sh and downloading a Blender tarball and running it), specifically https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh and https://download.blender.org/release/Blender4.2/blender-4.2.3-linux-x64.tar.xz, which fetch and execute external code needed for the skill to run.