The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/clean_temp.py performs file and directory deletion using os.remove and shutil.rmtree to free up disk space. This is the core functionality of the skill. To mitigate risk, the script uses a dry-run mode by default and requires the --execute flag followed by a manual 'YES' confirmation from the user before any files are actually removed.
[DATA_EXPOSURE]: The analysis scripts (analyze_disk.py and find_large_files.py) recursively scan the filesystem to identify large files and directories. This involves reading file metadata, names, and paths across system-wide locations including user application data and system logs.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads and displays file names from the local disk to the agent. If a file is maliciously named with LLM instructions, it could potentially influence the agent's behavior when the agent processes the scanning results. Evidence: The find_large_files.py and analyze_disk.py scripts ingest untrusted file names and output them directly to the console for the agent to read, without specific sanitization or boundary markers.

Geek-skills-c-drive-cleaner