The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's primary mechanism involves the dynamic generation and execution of JavaScript code. The run.js utility receives code as an argument or from a file, wraps it in an execution template, writes it to a temporary file in the skill directory (.temp-execution-*.js), and then executes it using the Node.js require() function. This allows for arbitrary code execution within the agent's runtime environment.
[COMMAND_EXECUTION]: The skill uses shell commands in run.js to manage the environment, including npm install and npx playwright install chromium via child_process.execSync. Additionally, the instructions in SKILL.md prompt the agent to execute node -e one-liners to interface with internal helper libraries.
[EXTERNAL_DOWNLOADS]: The skill automates the download of the Playwright library and Chromium browser binaries from official Microsoft/NPM registries during the setup phase and execution.
[DATA_EXFILTRATION]: The lib/helpers.js file contains a detectDevServers function that probes common network ports (3000, 3001, 8080, etc.) on localhost. While intended for developer server discovery, this constitutes local network reconnaissance.
[PROMPT_INJECTION]: As a tool that navigates and reads external web content, the skill is inherently susceptible to indirect prompt injection, where malicious instructions on a website could influence the agent's subsequent actions. The skill does not implement specific boundary markers or sanitization for website content before processing it.

playwright-skill