The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the mcp__claude-in-chrome__javascript_tool to execute arbitrary JavaScript code within the browser's page context.
Evidence: Multiple examples demonstrate using javascript_exec to interact with the DOM and window objects.
[DATA_EXFILTRATION]: The skill provides tools to monitor and capture potentially sensitive data from the browser environment.
Evidence: mcp__claude-in-chrome__read_network_requests captures all HTTP traffic, while mcp__claude-in-chrome__javascript_tool is explicitly shown being used to retrieve an authToken from localStorage.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted content from external web pages and possesses high-privilege interaction capabilities.
Ingestion points: read_page, read_console_messages, read_network_requests, and get_page_text (found in SKILL.md).
Boundary markers: None present to distinguish between instructions and data.
Capability inventory: javascript_tool (JS execution), computer (mouse/keyboard automation), and gif_creator (file export) (found in SKILL.md).
Sanitization: No sanitization or validation of ingested web content is described.

dev-test-chrome