The Agent Skills Directory

[COMMAND_EXECUTION]: Provides numerous shell commands and Node.js script templates for browser automation, screenshot capture, and visual regression analysis. These include npx serve, npx playwright install, and various script snippets for handling persistent browser sessions and element-specific captures.\n- [EXTERNAL_DOWNLOADS]: Instructions recommend installing standard Node.js packages such as playwright, pixelmatch, pngjs, serve, @11ty/eleventy, and wait-on from official registries.\n- [REMOTE_CODE_EXECUTION]: Uses node -e and npx to execute browser automation logic. The provided templates allow the agent to interact with locally running web applications and capture page data.\n- [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by processing external data from web pages. \n
Ingestion points: Page structure and content are ingested via page.accessibility.snapshot() and page.evaluate() (DOM serialization and element maps).\n
Boundary markers: Absent; the templates do not include specific delimiters or instructions for the agent to disregard instructions found within the processed web content.\n
Capability inventory: The skill allows for file writing (fs.writeFileSync), shell command execution (node -e), and network navigation (page.goto).\n
Sanitization: Some patterns include logic to remove <script> and event handler attributes from serialized HTML snapshots to reduce risk.

localhost-screenshots