The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions direct the agent to download and execute the @_davideast/stitch-mcp package directly from the NPM registry using npx or global installation.
[REMOTE_CODE_EXECUTION]: The init command is designed to download and install an isolated version of the Google Cloud SDK if it is missing from the system, which involves executing remote installation scripts.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and processes HTML and CSS code generated by Google's Stitch platform, creating an attack surface where malicious instructions embedded in designs could influence the agent.
Ingestion points: The get_screen_code, build_site, and serve commands fetch external design data from the Stitch API and load it into the agent's context or the local file system.
Boundary markers: There are no specified delimiters or instructions to ignore potential commands embedded within the retrieved HTML/CSS code.
Capability inventory: The skill possesses capabilities for network operations (Stitch API calls), file system modifications (Astro site generation), and local server execution (Vite dev server).
Sanitization: The instructions do not mention any sanitization or validation of the fetched design code before it is processed or served.
[COMMAND_EXECUTION]: The skill requires the execution of multiple shell commands to perform initialization, diagnostics, local serving, and site building.

stitch-mcp-cli