The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on several shell utilities to perform its core functions, including ls, find, du, mkdir, and mv. Specifically, it executes find with the -exec flag to run file and md5 on user files, which involves processing a large number of filesystem objects and executing subprocesses for each.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks (Category 8) because it ingests untrusted metadata and file content to determine how to organize the system.
Ingestion points: File names and types are ingested through ls, find, and the file utility. The instructions also suggest making decisions based on file 'content' (SKILL.md).
Boundary markers: There are no specified delimiters or 'ignore' instructions used when the agent processes file names or content, which could allow a maliciously named file to influence agent behavior.
Capability inventory: The skill has the capability to create directories (mkdir), move files (mv), and delete files (as indicated by the 'Delete' and 'Cleanup' tasks).
Sanitization: No sanitization of file paths or names is mentioned before they are passed into command strings or organization plans.

file-organizer