The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands to execute its Node.js scripts (shoot.mjs, contact-sheet.mjs) and to manage local development servers. It includes a specific command to terminate processes listening on a user-provided port (kill $(lsof -tiTCP:<port> -sTCP:LISTEN)), which is a standard developer workflow for cleaning up background tasks.
[DYNAMIC_EXECUTION]: The main script (shoot.mjs) dynamically imports a JavaScript configuration file (shots.mjs) to determine screenshot parameters and execution logic. This file can contain functions (setup, beforeShot, init, routes) that are executed within the Node.js process to manipulate the browser state. This is a core architectural feature of the skill for defining complex UI states.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local web applications by reading UI text, accessibility labels, and HTML source code via Playwright to identify targets for screenshots.
Ingestion points: Web application UI content and source code are read from local URLs (e.g., http://127.0.0.1:<port>) in scripts/shoot.mjs.
Boundary markers: No explicit boundary markers are used for the content of the screenshots, though the skill instructs the agent to manually review images for incorrect states.
Capability inventory: The skill has the capability to execute Node.js scripts, perform shell operations, and write image files to the local file system.
Sanitization: No automated sanitization is applied to the UI text being processed; however, the skill provides explicit instructions to use demo/fixture data and avoid real user content.

clean-ui-shot