higgsfield-shotlist-director
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data such as scripts, briefs, and treatments to generate a connected shotlist. This data is interpolated into prompt copy-blocks and an HTML artifact, which creates a potential surface for indirect prompt injection where malicious instructions in the input could influence the agent's behavior or the generated output.
- Ingestion points: User-provided brief, script, scene breakdown, or treatment processed at runtime.
- Boundary markers: The skill does not define specific delimiters or 'ignore' instructions for the untrusted data when interpolating it into the final HTML or prompt blocks.
- Capability inventory: The skill performs file writes (generating
shotlist.html) and executes a local Python script (seedance_lint.py) to validate prompts. - Sanitization: The instructions do not specify escaping or sanitization for user-provided text before it is placed into the HTML skeleton or prompt blocks.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script for prompt linting:
python3 scripts/seedance_lint.py --preflight --regime block --model seedance_2_0 "<prompt>". This is a standard part of the intended workflow within the repository's structure and targets local scripts rather than remote code.
Audit Metadata