The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to fetch and run the @playwright/cli package from the official NPM registry, ensuring the tool is available even if not installed globally.
[COMMAND_EXECUTION]: Provides a shell wrapper script scripts/playwright_cli.sh that mediates interaction between the agent and the Playwright CLI tool.
[REMOTE_CODE_EXECUTION]: Supports browser-level code execution via eval and run-code commands, allowing the agent to run arbitrary JavaScript within the target website's context for tasks like data extraction or complex interactions.
[PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it processes untrusted data from the web.
Ingestion points: The agent ingests external data through browser snapshots and DOM evaluations performed via pwcli open, pwcli snapshot, and pwcli eval in SKILL.md and references/cli.md.
Boundary markers: There are no explicit instructions or delimiters used to warn the agent to ignore instructions embedded within the processed web content.
Capability inventory: The agent possesses extensive capabilities to interact with the environment, including filling forms, clicking elements, and executing JavaScript code based on its interpretation of the page content.
Sanitization: No evidence of sanitization or filtering of the HTML/DOM content is present before the data is analyzed by the agent.

playwright