The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard system commands such as ls, find, du, mkdir, and mv to analyze and reorganize the file system. These operations are restricted to target directories defined by the user and are essential for the skill's primary function.
[SAFE]: The skill defines a structured workflow that emphasizes user consent, requiring the agent to present an 'Organization Plan' and obtain explicit confirmation before creating folders, moving files, or deleting duplicates. This human-in-the-loop requirement effectively mitigates risks associated with automated file management.
[PROMPT_INJECTION]: Potential surface for indirect prompt injection via file metadata during organization tasks. Evidence Chain: 1. Ingestion points: File names and types via ls and find commands documented in SKILL.md. 2. Boundary markers: Absent in command execution strings. 3. Capability inventory: File system modification via mkdir, mv, and rm (for duplicates) in SKILL.md. 4. Sanitization: None. This finding is noted for surface awareness but is considered safe due to the mandatory human-in-the-loop verification steps.

file-organizer