The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes Python scripts (init_skill.py, package_skill.py) that perform local file system operations such as directory creation and file writing. The init_skill.py script lacks strict input validation on the skill name before using it in path construction, which could lead to directory creation outside the intended path if supplied with traversal sequences.- [PRIVILEGE_ESCALATION]: The init_skill.py script automatically modifies file permissions using example_script.chmod(0o755). This makes a dynamically generated boilerplate script executable immediately upon creation.- [INDIRECT_PROMPT_INJECTION]: The skill generates new SKILL.md and script files by interpolating user-provided metadata, creating a surface for injection through data processing.
Ingestion points: User-supplied skill names and descriptions in init_skill.py flow directly into generated file headers and content.
Boundary markers: The generated templates lack clear delimiters or instructions to prevent the agent from obeying malicious instructions embedded in user-provided metadata.
Capability inventory: The skill possesses capabilities to write to the file system and modify file permissions via the provided Python utility scripts.
Sanitization: There is no evidence of input sanitization or escaping when incorporating user-provided text into the generated skill files.

skill-creator