The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on system commands to discover disk usage and perform cleanup tasks.\n
The scripts/scan.py utility uses subprocess.run to execute commands such as uname, brew, and docker to gather system information.\n
The agent is instructed to execute cleanup commands like rm, git branch -d, and osascript (for managing the Trash), all of which are governed by a strict requirement for explicit user approval before execution.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its core functionality of scanning the local filesystem.\n
Ingestion points: scripts/scan.py performs directory traversal and reads names of files and directories (e.g., within ~/Developer and ~/Downloads) to identify cleanup targets.\n
Capability inventory: The skill has the capability to perform file deletions and execute shell commands based on the paths it identifies.\n
Sanitization: The risk is mitigated by explicit instructions requiring the agent to present all findings to the user and obtain confirmation before running any command.\n
Boundary markers: The agent's instructions include clear rules to avoid system-critical directories and respect user-defined skip paths.

mac-cleanup