higgsfield-studio-router
Pass
Audited by Gen Agent Trust Hub on May 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted data from external catalog tools to drive generation parameters, which creates a surface for indirect prompt injection.
- Ingestion points: External catalog data ingested via
mcp__higgsfield__models_explore,mcp__higgsfield__show_marketing_studio, andmcp__higgsfield__show_characters. - Boundary markers: Absent. The skill does not instruct the agent to use delimiters or ignore instructions embedded in the catalog outputs.
- Capability inventory: The agent utilizes
mcp__higgsfield__generate_videoandmcp__higgsfield__generate_imageto execute content creation based on the ingested data. - Sanitization: Absent. There are no instructions to validate or escape strings returned from the discovery phase tools before they are used in generation prompts.
- [SAFE]: The skill contains several business-logic constraints and brand safety rules (e.g., 'NEVER mention SuperSeller', 'NEVER mention Marco') which act as internal safety filters for the agent's output.
- [SAFE]: No hardcoded credentials, sensitive file access patterns, or malicious external network operations were detected. All identified UUIDs and IDs appear to be platform-specific resource identifiers for public or semi-public avatars and presets.
Audit Metadata