render-model-comparison-grid
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The
scripts/render_seekable_hyperframe.pyfile executes theffmpegutility viasubprocess.run. The command construction utilizes a list of arguments rather than a shell string, which is a secure implementation that prevents shell command injection. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to how it handles configuration data during the rendering process.
- Ingestion points: Untrusted data (prompts, tags, and labels) is read from
config.jsoninscripts/build_composition.pyand interpolated into a generated HTML file. - Boundary markers: There are no explicit delimiters or boundary markers used in the HTML template to isolate user-provided text from the surrounding DOM structure.
- Capability inventory: The skill relies on Playwright for browser execution and
subprocess.runfor system-level command execution. - Sanitization: The scripts do not perform HTML escaping or sanitization when inserting data into the HTML via
innerHTML, which could allow for script execution within the local headless browser context during the rendering phase.
Audit Metadata