The Agent Skills Directory

[COMMAND_EXECUTION]: The provided Python scripts (init_skill.py, package_skill.py) perform sensitive file system operations including directory creation and file writing based on user-supplied arguments.\n- [COMMAND_EXECUTION]: A path traversal vulnerability exists in scripts/init_skill.py where the skill_name parameter is joined to the target path without validation. This allows a user to specify absolute paths or use parent directory references (e.g., ..) to create directories and write files outside of the intended destination.\n- [PRIVILEGE_ESCALATION]: The init_skill.py script programmatically modifies file permissions, using chmod(0o755) to grant executable status to generated script files.\n- [DYNAMIC_EXECUTION]: The skill includes functionality to generate and write new executable Python scripts from predefined templates at runtime.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an injection surface through its reliance on user-supplied arguments which are processed by scripts with significant system capabilities.\n
Ingestion points: The skill_name and path arguments provided to scripts/init_skill.py and scripts/package_skill.py via sys.argv.\n
Boundary markers: None present; user-supplied strings are interpolated directly into file paths and template content.\n
Capability inventory: File system write access (Path.write_text), directory creation (Path.mkdir), permission modification (Path.chmod), and file archival (zipfile.ZipFile).\n
Sanitization: Absent; the script prints naming requirements in its help message but does not programmatically enforce them before performing file operations.

skill-creator