The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run within render_artifact.py and test_skill.py to automate its internal validation and rendering pipeline.
This execution is strictly controlled, targeting the skill's own scripts (e.g., preflight_artifact.py) using the current Python interpreter (sys.executable).
It serves as a security layer by ensuring that pre-rendering checks are automatically enforced before the Playwright-based renderer is initialized.
[EXTERNAL_DOWNLOADS]: The skill utilizes the playwright package, which requires a one-time download of a browser binary (playwright install chromium). This is a standard and expected operation for a tool whose primary purpose is to render complex HTML/CSS layouts into static assets.
[SAFE]: The rendering implementation in render_artifact.py follows security best practices for handling potentially untrusted HTML content.
It explicitly disables JavaScript execution (java_script_enabled=False) and blocks all outgoing HTTP/HTTPS network requests by default.
These protections ensure that the agent remains in a local, offline environment during the visual production phase unless the user manually grants overrides for specific needs.
[SAFE]: The inclusion of preflight_artifact.py provides an automated 'Anti-Slop' and security audit of the generated source code. It scans for:
Unresolved template placeholders (e.g., [[HEADLINE]]).
Embedded scripts that might attempt to execute code.
External resource links that would break in an offline render or leak metadata.
[SAFE]: The skill uses skill_view to resolve its own directory path, ensuring that file operations are relative to the skill's installation directory rather than relying on an unpredictable current working directory.

marketing-collateral-design