The Agent Skills Directory

[COMMAND_EXECUTION]: The skill installs OS-level persistence to run background tasks automatically. This includes creating hourly scheduled tasks via schtasks on Windows, adding LaunchAgents plist files on macOS, and modifying crontab on Linux.
[COMMAND_EXECUTION]: The setup script modifies the internal configuration of the Claude Code agent (~/.claude/settings.json) to inject a custom 'Stop' hook that executes the skill's monitoring script every time the agent finishes a session.
[DATA_EXFILTRATION]: The skill collects and transmits sensitive AI usage telemetry and device identity information to an external endpoint (https://api.agenticcommons.xyz/v1/usage/daily). The data includes token counts, model names, and system metadata such as hostname, CPU profile, and OS version.
[DATA_EXFILTRATION]: The collection scripts automatically read and parse private local data sources, including Claude Code statistics, Codex session JSONL logs, Gemini session files, and OpenCode SQLite databases, to extract usage metadata.
[COMMAND_EXECUTION]: The skill utilizes spawn to execute PowerShell command strings on Windows systems for the purpose of decrypting saved access tokens.

acommons