The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the playwright-cli binary to perform browser-based operations via the system shell.\n- [REMOTE_CODE_EXECUTION]: The run-code command allows the agent to execute arbitrary JavaScript/Playwright code within the browser context, which can be used to perform complex logic that bypasses standard CLI command limitations. Evidence: playwright-cli run-code "async page => { ... }" in references/running-code.md.\n- [DATA_EXFILTRATION]: The skill includes extensive functionality to retrieve sensitive session data, such as cookies, local storage, and clipboard content. Evidence: playwright-cli cookie-list and playwright-cli state-save auth.json in references/storage-state.md, and clipboard access examples in references/running-code.md.\n- [PROMPT_INJECTION]: The skill is highly vulnerable to indirect prompt injection as it ingests untrusted content from web pages.\n
Ingestion points: playwright-cli snapshot, playwright-cli eval, and the browser DOM accessed via run-code.\n
Boundary markers: Absent; there are no instructions for the agent to treat website content as untrusted or to ignore embedded instructions.\n
Capability inventory: Shell command execution (playwright-cli), file system writes (state-save, screenshot), and full network access (inherent to browser automation).\n
Sanitization: Absent; the skill does not implement or describe methods for sanitizing data retrieved from the web before processing.\n- [EXTERNAL_DOWNLOADS]: The skill provides commands to download and install browser binaries and external skill components. Evidence: playwright-cli install-browser and playwright-cli install --skills in SKILL.md.

playwright-cli