The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill is designed to process untrusted external content (repomix-packed files) and convert it into local files and directories. This is a classic surface for Path Traversal attacks.
Ingestion points: The unmix_repomix.py script parses paths and content from user-provided files (XML <file path="...">, Markdown ## File: ..., or JSON "path": "...").
Boundary markers: The skill uses regex and standard format parsers to identify file boundaries, which can be bypassed or manipulated by malformed input.
Capability inventory: The skill explicitly uses python3 to execute a script that has the capability to create directories and write arbitrary file content to the filesystem.
Sanitization: There is no mention of path sanitization or validation in the documentation. Without strict checks (e.g., preventing .. or absolute paths), an attacker could overwrite ~/.ssh/authorized_keys, .bashrc, or other critical files.
[Command Execution] (MEDIUM): The skill relies on executing a local Python script (scripts/unmix_repomix.py) via the CLI. While intended for functionality, it serves as the execution engine for the risks identified in the file-writing process.

repomix-unmixer