media-higgsfield-identity
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust decision-making workflow to guide the AI agent in selecting between two distinct technical paths (Soul vs Element) for maintaining visual consistency.
- [SAFE]: References to external sources (github.com/higgsfield-ai/skills) point to official documentation for the underlying service.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided media and names which are then interpolated into prompts for image/video generation tools (e.g., using
<<<element_id>>>). While this represents a data ingestion surface, the risk is low as it follows the expected tool-chaining behavior of the platform. Evidence: Ingestion points in SKILL.md and references/soul-vs-elements.md; Capability inventory includes show_characters, show_reference_elements, generate_image, and generate_video; Boundary markers are present in the form of specific ID encapsulation (<<>>). - [SAFE]: The skill includes explicit warnings regarding portrait rights and user responsibility, promoting ethical use of identity training.
Audit Metadata