The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves interacting with external web content, which serves as an attack surface for indirect prompt injection. Malicious instructions embedded in websites could potentially influence the agent's behavior during automation tasks.
Ingestion points: The skill utilizes page.goto(url) to access external sites and methods like textContent() to extract data, as seen in the examples in references/detailed-guide.md.
Boundary markers: The guide instructs treating CAPTCHAs, access denials, and locked desktops as explicit boundaries, though these are human-level rather than programmatic markers.
Capability inventory: The skill uses extensive browser automation capabilities, including page navigation, input manipulation, network request interception, and sensitive file generation (traces/auth state).
Sanitization: There is no mention of sanitizing or filtering the content extracted from web pages before it is processed by the agent.
[COMMAND_EXECUTION]: The documentation includes instructions for the agent or user to execute CLI commands to run tests and debug failures.
Evidence: The guide includes commands like npx playwright test and npx playwright show-trace for managing the automation workflow.

browser-automation