The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard macOS utilities and local Python scripts to discover bloat and perform cleanup operations.
Evidence:
scripts/scan_reclaimable.py executes system commands including du, find, git, and tmutil via subprocess.run to analyze disk usage.
scripts/safe_delete.py performs permanent file and directory removal using shutil.rmtree and path.unlink once a path is validated.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface by ingesting and processing untrusted directory and file names from the user's local filesystem (Category 8).
Ingestion points: File and directory paths are read from the $HOME directory by scripts/scan_reclaimable.py during its discovery sweep.
Boundary markers: The skill does not use specific delimiters or instructions to prevent the agent from interpreting file names as commands, though the structured output and script-based deletion mitigate this risk.
Capability inventory: The agent can permanently delete files and execute tool-specific cleanup commands (e.g., git gc, brew cleanup).
Sanitization: scripts/safe_delete.py implements comprehensive sanitization by checking paths against an extensive DENY_PREFIXES list (protecting sensitive user data and system volumes) and validating them against a PURGE_DIR_NAMES allowlist.

mac-disk-reclaim