The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted JSON configuration data to generate HTML mockups. Although it uses an escapeHTML function to sanitize text fields, the overall architecture presents a surface for indirect prompt injection where malicious instructions could be embedded in data fields.
Ingestion points: JSON configuration files loaded via render.js.
Boundary markers: An escapeHTML function is used to delimit text content in generate.js.
Capability inventory: The skill can read local files, access the network for assets, and render content in a headless browser via Playwright.
Sanitization: Character escaping is applied to most text fields, but file paths and URLs are used directly for resource loading.
[EXTERNAL_DOWNLOADS]: The skill fetches typography from Google Fonts (fonts.googleapis.com and fonts.gstatic.com) and placeholder images from i.pravatar.cc to populate the rendered mockups.
[REMOTE_CODE_EXECUTION]: The installation process requires downloading the Playwright library and the Chromium browser engine (npx playwright install chromium) which are executed locally to capture screenshots.
[DATA_EXFILTRATION]: The skill resolves and loads local files specified in the configuration for wallpapers and icons. Because it uses the file:// protocol without a restricted allowlist, it could be coerced into rendering sensitive local files (e.g., configuration or credential files) into the output screenshot.
[COMMAND_EXECUTION]: The documentation provides instructions for executing the renderer and tests using CLI commands (node render.js and bash tests/run-all.sh).

render-ios-lockscreen