The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides boilerplate JavaScript code that uses the Playwright library to perform automated UI actions like clicking, typing, and navigating. These scripts are intended to run in a local environment to facilitate UI testing and debugging.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where the agent processes content from web applications (URLs, page titles, DOM elements). Ingestion points: Page content retrieved via page.goto, page.title(), and page.url() in SKILL.md. Boundary markers: Absent; the skill relies on stable app markers for selection but does not use explicit delimiters for page-derived text. Capability inventory: Browser automation, screenshot capture, and local file writing to /tmp/ across all code snippets in SKILL.md. Sanitization: No explicit sanitization or filtering of content retrieved from the browser is described.
[DYNAMIC_EXECUTION]: The skill facilitates the generation of diagnostic scripts from templates to interact with local dev servers or Chromium debug ports. This includes capabilities for performance profiling, heap snapshots, and network inspection.

control-ui