higgsfield-pipeline
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user input, such as project concepts, master scripts, and scene descriptions, to generate structured prompts for video generation models. This ingestion of external data creates a surface for indirect prompt injection if the user content contains instructions intended to override agent behavior during the prompt generation or assembly stages.\n
- Ingestion points: User-provided project ideas, Master Scripts, and plain-language scene descriptions (SKILL.md, Step 03, Pipeline E Stage 4).\n
- Boundary markers: The skill recommends structured formatting (Identity vs. Motion blocks, Prompt Modules) which provides logical separation but does not use explicit boundary delimiters to isolate user input from agent instructions.\n
- Capability inventory: The agent generates complex text prompts and provides instructions for interacting with the Higgsfield platform's video generation and editing tools.\n
- Sanitization: The skill does not describe any methods for sanitizing, escaping, or validating the user-provided content before it is interpolated into prompt generation workflows.
Audit Metadata