The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to clone a complete full-stack application from an unverified third-party repository at https://github.com/hoangsonww/Claude-Code-Agent-Monitor.git. This repository is outside the author's stated infrastructure.
[REMOTE_CODE_EXECUTION]: The installation process involves running npm install and npm start on the downloaded external code, which executes arbitrary third-party scripts on the host system.
[COMMAND_EXECUTION]: The skill uses child_process.spawn within the Electron desktop application to launch backend services. It also directs users to use chmod +x to grant execution permissions to a downloaded Python hook script.
[DATA_EXFILTRATION]: The skill provides a Python hook script (hook_handler.py) designed to be integrated into the Claude Code agent. This script intercepts all tool usage events and session data, including sensitive metadata such as file paths and durations, and transmits them to a configurable external DASHBOARD_URL via HTTP POST requests.
[COMMAND_EXECUTION]: The skill establishes persistence by modifying the agent's global configuration file (~/.claude-code/config.json) to execute the monitoring script for every tool invocation. Additionally, the desktop applications are configured to use SMAppService or the startup folder to ensure they launch automatically upon user login.

claude-code-agent-monitor