The Agent Skills Directory

[COMMAND_EXECUTION]: The skill exposes the run-code and eval commands, which enable the execution of arbitrary JavaScript code within the browser automation context. This capability allows for complex interactions, including file system operations such as download.saveAs().
[DATA_EXFILTRATION]: The skill provides tools for accessing and exporting sensitive browser state, including cookies, local storage, and session storage via commands like state-save, cookie-list, and localstorage-get. These can be used to extract authentication tokens or active session identifiers from the browser environment.
[EXTERNAL_DOWNLOADS]: The installation instructions reference downloading the @playwright/cli package from the public NPM registry. This is a standard procedure for browser automation tools.
[PROMPT_INJECTION]: The skill processes untrusted content from external websites through navigation and snapshots. This creates an attack surface for indirect prompt injection, where a malicious site could influence the agent's behavior by embedding instructions in the DOM.
Ingestion points: playwright-cli goto, playwright-cli snapshot, and playwright-cli eval in SKILL.md and references/element-attributes.md.
Boundary markers: No explicit markers are used to isolate ingested website content from the agent's instructions.
Capability inventory: High; includes shell command execution via tools, file system writes, and arbitrary JavaScript execution in the browser context.
Sanitization: No sanitization of ingested HTML or text is performed before it is presented to the agent for analysis.

playwright-cli