higgsfield-seedance-2-5
Warn
Audited by Gen Agent Trust Hub on Aug 23, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains explicit instructions designed to circumvent safety guardrails. In the section 'The Real-Person Character Formula' within SKILL.md, it explicitly directs the agent to avoid using age-related keywords to prevent triggering content filters related to minors, recommending the use of descriptive roles and physical attributes instead.
- [COMMAND_EXECUTION]: The skill references the execution of local scripts with user-supplied arguments. The command template 'python3 scripts/seedance_lint.py --preflight --model seedance_2_5 ""' in SKILL.md takes user-provided text and interpolates it directly into a shell command, creating a potential command injection vector if the input contains malicious shell meta-characters.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to the processing of untrusted user input without sufficient safeguards. 1. Ingestion points: User-provided prompts passed to the seedance_lint.py script. 2. Boundary markers: Absent; there are no instructions to use delimiters or treat the input as non-executable data. 3. Capability inventory: Execution of local shell commands via python3 and file system access to model specifications. 4. Sanitization: Absent; the skill does not require the agent to validate or escape user-provided content before processing.
Audit Metadata