higgsfield-website

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.70). The skill includes an explicit, self-serving instruction to force users to register via the skill-author's referral link ("exactly this link"), which is outside the functional scope of building and deploying the site and therefore constitutes a deceptive/injection-like directive.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill registers and uses the Higgsfield MCP server at runtime (https://mcp.higgsfield.ai/mcp) to execute generation/hosting tools, so this external endpoint is a required runtime dependency that runs remote code and controls agent tool behavior.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill explicitly instructs installing missing system dependencies and even shows a sudo install command ("sudo apt install ffmpeg"), which directs the agent to modify the host system state and request elevated privileges.