The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates for standard shell utilities including ls, find, du, mkdir, and mv. These commands are used to analyze file types, identify duplicates via MD5 hashing, and reorganize directory structures, which are all appropriate activities for the stated purpose of file management.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface by reading file names and directory structures using ls and find. While these inputs are untrusted and could theoretically contain malicious instructions, the skill defines an "Execution Workflow" and "Important Rules" requiring the agent to confirm all moves and deletions with the user. The presence of these human-in-the-loop checkpoints mitigates the risk of the agent automatically following instructions embedded in file metadata.
Ingestion points: File names and directory listings via ls and find commands (SKILL.md).
Boundary markers: The skill uses markdown code blocks to separate instructions from command templates but lacks specific "ignore instructions" delimiters for file data.
Capability inventory: File system modification via mkdir and mv, and command execution through the find -exec pattern.
Sanitization: No specific sanitization or escaping of file names is mentioned in the provided templates.

file-organization