The Agent Skills Directory

[SAFE]: The skill architecture forces a secure multi-phase workflow (scan, analyze, preview, dry-run) before any filesystem modifications occur, ensuring that the user has full visibility into proposed changes.
[SAFE]: The project_organizer.py script implements an extensive PROTECTED_PATTERNS list that inherently prevents the modification of sensitive data, including version control directories (.git, .svn), environment secrets (.env, credentials), and lock files.
[SAFE]: No network access, external telemetry, or remote code execution patterns were identified. The implementation relies exclusively on Python standard libraries for local operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted filesystem data (file and directory names). 1. Ingestion point: scan_project function in scripts/project_organizer.py scans a user-defined directory. 2. Boundary markers: The skill mandates a 5-step validation workflow and explicit 'yes' confirmation at the CLI prompt before execution. 3. Capability inventory: Move files, delete empty directories, and create organizational folders. 4. Sanitization: The is_protected method validates all targets against a safety list, and delete operations are restricted to confirmed empty directories. Due to these strong controls, the potential for malicious injection via filenames is effectively mitigated.

safe-project-organizer