The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and run the playwriter package directly from the npm registry using npx playwriter@latest or bunx playwriter@latest.
[COMMAND_EXECUTION]: Local shell commands are used to invoke the playwriter CLI tool to interact with the browser environment.
[REMOTE_CODE_EXECUTION]: The skill provides the ability to execute arbitrary JavaScript code snippets within the browser's context using the -e flag. Additionally, the instruction to run playwriter skill to fetch documentation at runtime constitutes dynamic loading of logic that is not visible in the provided skill files.
[DATA_EXFILTRATION]: The skill connects to the user's existing Chrome browser instance rather than a fresh session. This gives the agent access to active authenticated sessions, cookies, and private history on sensitive sites such as Instagram and Twitter, posing a risk of session theft or data exposure.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it is designed to ingest and process content from external, third-party websites which may contain malicious instructions targeting the agent's browser control capabilities.
Ingestion points: Content from external websites loaded via page.goto inside the Playwright execution strings.
Boundary markers: Absent; there are no instructions to delineate or treat external web content as untrusted.
Capability inventory: Shell command execution and arbitrary browser-level JavaScript execution.
Sanitization: None described; the skill does not specify any methods for validating data retrieved from the web.

playwriter