The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data by scanning directory contents, including filenames and file types. Maliciously crafted filenames could be used to attempt to influence the agent's decision-making process or bypass instructions when the file list is processed and displayed back to the user.
Ingestion points: Target directories and their contents (filenames, metadata).
Boundary markers: The skill does not explicitly define delimiters to separate file metadata from agent instructions.
Capability inventory: The skill uses mv, mkdir, and find to manipulate the file system.
Sanitization: While the skill uses quoting in its command templates, it does not explicitly mention sanitizing special characters in filenames before processing.
[COMMAND_EXECUTION]: The skill executes shell commands (find, du, mv, mkdir) to analyze and restructure the file system. While the instructions include best practices such as requesting user confirmation before execution and logging actions, these capabilities allow the agent to perform broad modifications to the user's data.

file-organizer