marketing-collateral-design
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes internal scripts such as preflight_artifact.py using subprocess.run. These calls use structured argument lists and are limited to the skill's own utility scripts.
- [EXTERNAL_DOWNLOADS]: The skill manages dependencies via pip and installs the Chromium browser using Playwright. These downloads are performed from well-known registries and trusted organizations.
- [DYNAMIC_EXECUTION]: The skill renders artboards using Playwright. Potential risks are mitigated by default settings that disable JavaScript and block all HTTP/HTTPS network traffic during rendering.
- [INDIRECT_PROMPT_INJECTION]: The skill processes HTML and CSS templates which could contain instructions from external sources.
- Ingestion points: scripts/render_artifact.py and scripts/preflight_artifact.py ingest HTML files.
- Boundary markers: The preflight script identifies and flags unresolved placeholders.
- Capability inventory: The toolchain involves file writing and internal script execution via subprocess.
- Sanitization: Robust sanitization is implemented through static analysis and sandboxed browser execution with JS and network access disabled by default.
Audit Metadata