The Agent Skills Directory

[DYNAMIC_EXECUTION]: The run.js script dynamically assembles and executes JavaScript code at runtime. It takes input from file paths, command-line arguments, or stdin, wraps it in an async IIFE template, writes it to a temporary file (.temp-execution-*.js), and then executes it using require(tempFile). This pattern allows the agent to execute arbitrary code, which could be exploited if malicious input is provided to the agent.
[REMOTE_CODE_EXECUTION]: The skill automatically manages its environment by executing shell commands to install software. The run.js script uses execSync to run npm install and npx playwright install chromium if dependencies are missing. While targeting the official Playwright registry, this involves downloading and executing code from external sources during the skill's operation.
[INDIRECT_PROMPT_INJECTION]: As a browser automation tool, the skill is designed to ingest and process content from external websites via Playwright. This represents a significant attack surface where a malicious website could embed instructions designed to manipulate the agent's behavior when it parses the page content.
Ingestion points: Browser page content is read via extractTexts, extractTableData, and page.evaluate in lib/helpers.js.
Boundary markers: The code does not implement explicit delimiters or instruction-bypass guards when processing extracted text.
Capability inventory: The skill has access to the filesystem (writing temp files), network (browser navigation), and local command execution (execSync, require).
Sanitization: There is no evidence of sanitization or filtering of the content extracted from web pages before it is returned to the agent.
[COMMAND_EXECUTION]: The skill invokes system-level commands through execSync in run.js and via the scripts section in package.json to manage package installation and browser binaries.

playwright-v1