The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external web pages during its discovery phase, creating a potential vector for malicious instructions to influence the agent.
Ingestion points: The Phase 1 Discovery script uses page.evaluate to scrape DOM attributes including tagName, type, name, placeholder, and textContent from interactive elements.
Boundary markers: The extracted data is not delimited or qualified with instructions to ignore potential commands embedded within the page content.
Capability inventory: The skill utilizes the Playwright library to automate browser interactions, performs file system operations via fs.copyFileSync, and executes shell commands to run the generated recording scripts.
Sanitization: There is no evidence of sanitization, filtering, or escaping applied to the scraped DOM content before it is processed by the agent to create the recording script.
[DYNAMIC_EXECUTION]: The skill's workflow depends on the generation of a JavaScript file (demo-script.cjs) incorporating locally defined logic and execution of that file via the Node.js runtime.

ui-demo