The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill uses npx --yes @playwright/cli@latest to download and execute the Playwright CLI. This tool is part of a package scope (@playwright) managed by a trusted organization (Microsoft).
[DYNAMIC_EXECUTION]: The skill provides instructions for using run-code and eval commands which execute arbitrary JavaScript strings within the browser context. This functionality is intended for inspection and waiting tasks during automation.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external web environments, creating a potential path for indirect prompt injection.
Ingestion points: The agent ingests external data from web pages via open, goto, snapshot, find, console, and requests commands (SKILL.md, references/mcp-to-cli-map.md).
Boundary markers: There are no explicit instructions or delimiters defined to separate web content from agent instructions or to prevent the agent from obeying commands embedded in page data.
Capability inventory: The skill provides high-privilege UI interaction capabilities including run-code, eval, click, fill, type, and press (SKILL.md).
Sanitization: No specific content sanitization, filtering, or validation mechanisms for external web content are mentioned in the protocol.

protocol-browser-anti-stall