The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local command-line interface (tabbit-cli) to manage task-isolated browser workspaces. It provides explicit paths for macOS, Linux, and Windows installations and includes rigorous validation steps for identifying legitimate browser instances through registry and file system checks.
[DYNAMIC_EXECUTION]: The agent generates and executes JavaScript snippets using the Playwright framework within a persistent Node.js realm. These scripts are used for navigation, form submission, and data extraction, following established safety 'recipes' that emphasize bounding loops and returning compact, JSON-safe results.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external web content, which presents a surface for indirect injection attacks.
Ingestion points: Web page content, DOM elements, and ARIA snapshots accessed via Playwright.
Boundary markers: Instructions explicitly require bounding result sizes, deduplicating data, and verifying output consistency before returning information to the agent context.
Capability inventory: The skill allows for file writing to local artifact directories (artifactPath) and standard browser-based network interactions.
Sanitization: Guidelines enforce the extraction of specific fields and the use of semantic locators to minimize the ingestion of malicious or irrelevant web data.

tabbit