The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute chrome-devtools CLI commands for controlling browser instances.
It executes local shell commands using sed to read connection metadata (port and websocket path) from the DevToolsActivePort file located in the user's Chrome application support directory.
It uses ls to enumerate existing persistent browser profiles stored in $HOME/.cache/chrome-devtools-mcp/profiles.
[EXTERNAL_DOWNLOADS]: The documentation encourages users to install and maintain the latest version of the chrome-devtools-mcp package from the NPM registry.
[PROMPT_INJECTION]: The skill facilitates the ingestion of data from external, untrusted web environments, which constitutes an indirect prompt injection attack surface.
Ingestion points: Data enters the agent's context through tools like take_snapshot (accessibility tree), evaluate_script (DOM data), list_console_messages (browser logs), and list_network_requests (network traffic metadata).
Boundary markers: The skill instructions do not specify any delimiters or safety prompts to help the agent distinguish between its core instructions and potentially malicious content retrieved from the web.
Capability inventory: The agent possesses powerful capabilities including arbitrary shell command execution (Bash), local file access (Read), and comprehensive browser control (executing JavaScript, managing extensions, and uploading files).
Sanitization: There are no explicit mechanisms defined to sanitize, filter, or validate data retrieved from web pages before it is processed by the agent.

browser