The Agent Skills Directory

[COMMAND_EXECUTION]: The skill operates by executing shell commands through 'npx prima-cli' and 'npx playwright-cli' to drive browser sessions, generate reports, and manage browser instances.
[EXTERNAL_DOWNLOADS]: Uses 'npx' to fetch and run the 'prima-cli' and 'playwright-cli' packages from the npm registry. It also triggers 'npx playwright install' to download browser binaries from Microsoft's official sources.
[DYNAMIC_EXECUTION]: Provides a 'pw' command that accepts and executes arbitrary JavaScript code snippets within the browser context to interact with the Playwright 'page' object, serving as a fallback for the AI automation layer.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes live content from web pages, which creates a potential surface for malicious instructions to influence the agent's actions.
Ingestion points: Browser page content (HTML, ARIA trees) and screenshots are analyzed by the 'check', 'ask', and 'research' commands.
Boundary markers: The documentation does not specify the use of clear delimiters or boundary markers to separate untrusted web content from internal agent instructions.
Capability inventory: The skill has capabilities including shell command execution, browser manipulation, and writing session artifacts (HTML, screenshots) to the local disk.
Sanitization: There is no mention of specific sanitization or filtering protocols for external web data before it is processed by the AI models.

prima