The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The skill utilizes several powerful shell commands (find, mv, mkdir, du) to manage the filesystem. These commands are constructed dynamically using variables representing file paths and names, which is a significant security risk if the input is not sanitized.
[PROMPT_INJECTION] (HIGH): This skill exhibits a major Indirect Prompt Injection surface (Category 8). 1. Ingestion points: The skill reads untrusted data (file names, types, and paths) directly from the user's filesystem using ls -la and find (SKILL.md, Instruction 2). 2. Boundary markers: There are no delimiters or boundary markers used when processing these file names to distinguish data from instructions. 3. Capability inventory: The skill has significant write capabilities including creating directories and moving/renaming files (mkdir, mv) across the system (SKILL.md, Instruction 6). 4. Sanitization: The instructions lack any requirement to sanitize file names or paths. An attacker could place a file with a name containing shell metacharacters (e.g., ;, |, `) or escape sequences, which could cause the agent to execute unintended commands during the analysis or execution phases.

file-organizer