The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command to run Google Chrome in headless mode to take screenshots of the generated prototype. This involves launching a local binary with parameters and using the $PWD shell variable. Evidence found in Step 4: "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" --headless --screenshot=shots/<state>-<size>.png --window-size=<640 or 1280>,2400 "file://$PWD/prototype.html#<state>[-desktop]".
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data by reading local product source code to extract design tokens and components. This creates a surface where malicious instructions hidden in code comments or data strings within the source files could influence the agent's behavior during the build or screenshot phase. (1) Ingestion points: Local product source code files. (2) Boundary markers: No delimiters or ignore-instructions are specified for the ingested code content. (3) Capability inventory: File system reading, file system writing, and shell command execution. (4) Sanitization: The skill does not specify any sanitization or validation of the code content before processing.
[DYNAMIC_EXECUTION]: The skill performs dynamic generation of an HTML file containing inline CSS and JavaScript based on the analysis of existing components and design tokens in the source code.

html-prototype