The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a shell command using node -e to parse the JSON output of a Lighthouse audit stored in /tmp/lh-report/report.json.
[INDIRECT_PROMPT_INJECTION]: The skill possesses the capability to ingest untrusted data from external websites while having powerful tool access, which represents a potential attack surface.
Ingestion points: Network request headers, console messages, and page snapshots from external domains are retrieved using tools like get_network_request and list_console_messages.
Boundary markers: The instructions do not define specific delimiters for separating untrusted website data from the agent's instructions.
Capability inventory: The skill uses capabilities including evaluate_script, click, navigate_page, and lighthouse_audit via the Chrome DevTools MCP.
Sanitization: No explicit sanitization or filtering of external website-provided content is performed before the agent processes it.
[SAFE]: All external resources and tools described are part of the intended browser debugging functionality provided by the Chrome DevTools environment.

cookie-debugging