The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell commands for environment initialization and health monitoring.
Evidence: Provides the command wsh server -L agent-$$ --ephemeral & to start a local terminal server if one is not already running.
Evidence: Includes a curl command using a Unix domain socket (${XDG_RUNTIME_DIR:-/tmp}/wsh/default.http.sock) to perform a health check against a local endpoint.
[PROMPT_INJECTION]: The skill contains instructional guidance rather than attempts to override the underlying agent's core safety or instructions. It provides examples of what the agent should look for (e.g., "rm -rf", "chmod 777") to protect the user's environment.
[DATA_EXFILTRATION]: No network requests to external or untrusted domains were detected. All network-like operations (curl) are directed at the local machine via Unix sockets for tool-to-agent communication.
[EXTERNAL_DOWNLOADS]: The skill does not instruct the agent to download or execute external scripts or packages from the internet.
[SAFE]: The skill focuses on providing user-centric terminal assistance and explicitly advises against being disruptive or capturing private data unnecessarily.

monitor