The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/file_watcher.py uses subprocess.run with shell=True to execute arbitrary shell commands passed via the --exec argument. While functional for a file-watching utility, this provides a mechanism for executing code whenever file changes are detected.
[INDIRECT_PROMPT_INJECTION]: The skill monitors the file system for changes, creating an attack surface where external events (like file creation or modification) trigger automated shell command execution.
Ingestion points: The script monitors the local file system using os.walk and os.stat in scripts/file_watcher.py to detect changes.
Boundary markers: The script does not implement safety boundaries or user confirmation for each event-triggered execution; it runs the command automatically when the monitored state changes.
Capability inventory: Includes directory traversal (os.walk), file metadata retrieval (os.stat), and shell command execution (subprocess.run).
Sanitization: No validation or sanitization is performed on the trigger events or the command provided to the --exec argument.

file-watcher