higgsfield-shotlist-director
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local Python script
scripts/seedance_lint.pyto perform linting on generated prompts. The command incorporates the generated prompt as a command-line argument:python3 scripts/seedance_lint.py --preflight --regime block --model seedance_2_0 "<prompt>". - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted user data (scripts, briefs, scene breakdowns) and transform it into a shotlist. 1. Ingestion points: User-provided text content (briefs, scripts, etc.) processed to create scenes. 2. Boundary markers: None explicitly mentioned for the user input processing within the instructions. 3. Capability inventory: The skill can write files (
shotlist.html) and execute local shell commands (python3 scripts/seedance_lint.py). 4. Sanitization: A "preflight linter" is utilized to check for policy violations and formatting issues in the generated prompts. - [DYNAMIC_EXECUTION]: The skill generates a self-contained HTML file containing inline JavaScript for managing UI state (checkboxes) and user interactions (copying prompts to the clipboard). The generated content incorporates user-provided descriptions, which could potentially introduce cross-site scripting (XSS) risks if the input is not properly escaped before being placed into the HTML template.
Audit Metadata