The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches and executes the create-mastra project initializer and associated dependencies such as @mastra/stagehand and @mastra/agent-browser from official package registries.
[COMMAND_EXECUTION]: Executes shell commands to scaffold projects using various package managers including npm, yarn, pnpm, and bun, and manages the local development server via npm run dev.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied parameters (directory, name, tag) which are interpolated into shell commands.
Ingestion points: The ARGUMENTS string parsed during the initial validation step.
Boundary markers: Not explicitly present in the shell command templates, though the skill provides instructions for the agent to validate input.
Capability inventory: Shell execution (npx, npm run dev), file system writes, and browser automation via Chrome MCP tools.
Sanitization: The instructions direct the agent to validate that names contain no special characters and that parameters for package managers and LLM providers match a predefined whitelist.
[DYNAMIC_EXECUTION]: Dynamically generates a TypeScript file (browser-agent.ts) with configuration based on user-selected providers and executes this code through the development server environment.

smoke-test