The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/platforms.py script uses subprocess.run to execute pgrep (on macOS/Linux) and tasklist (on Windows). These calls are used exclusively to determine if specific applications are running to avoid file corruption. The arguments are sourced from a static internal map of process names, effectively preventing command injection.
[INDIRECT_PROMPT_INJECTION]: The skill ingests file metadata and filenames which could contain malicious strings. However, the skill implements a comprehensive security model including a denylist for sensitive directories (e.g., .ssh, .aws, .gnupg) and explicit instructions for the agent to treat all file content as data rather than instructions.
[DATA_EXFILTRATION]: All photo scanning and duplicate detection logic is performed locally using Python standard libraries. There are no network-capable functions (curl, requests, socket) present in the skill's scripts.
[PRIVILEGE_ESCALATION]: The skill uses platform-specific Trash APIs (such as SHFileOperationW on Windows) for file removals. This ensures that operations are performed within the user's standard permission context and remain reversible via the system's native Recycle Bin or Trash folder.

organizing-photos