higgsfield

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow ingests outsider-authored free text from the ${var} request body (user-supplied prompt/hints and, for animate:, a user-supplied source image URL) and passes it into the Higgsfield MCP generation tool via mcp__higgsfield__*, so attacker-provided text can directly reach the LLM at runtime.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill explicitly performs billable actions that consume the operator's Higgsfield account credits (irreversible spend). It requires OAuth connection to the operator's account, reports cost/credit figures, handles "no credits" / payment-required errors, and enforces spend limits per run — i.e., it issues API calls that debit the user's account for each generation. This is direct financial execution (spending account credits) rather than a generic tool.