higgsfield-seedance-prompt
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides purely instructional content for generating high-quality prompts for video AI. It does not include any executable code, shell commands, or network operations. Analysis of the following categories confirmed no risks:
- Data Exfiltration & Exposure: No hardcoded secrets, sensitive file paths, or network calls (curl, wget, fetch) were detected.
- Prompt Injection: The instructions do not contain safety filter bypasses, DAN-style overrides, or instructions to ignore previous rules. The phrase 'Always consult this skill before writing any Seedance prompt' is a legitimate operational instruction for the agent.
- Obfuscation: No Base64, hex encoding, zero-width characters, or homoglyph attacks were found in the text or metadata.
- Dependencies: The skill does not install or reference any external packages or scripts.
- Dynamic Execution: There are no instances of dynamic code generation or context injection patterns (such as the
!commandsyntax).
Audit Metadata