The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Write Bash tool and provides two Python scripts (init_skill.py and package_skill.py) to manage the file system. These scripts automate directory scaffolding and the creation of ZIP archives for skill distribution.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided inputs such as skill names and purpose descriptions to generate file content, which could be leveraged to inject instructions into new skill templates.
Ingestion points: User-supplied <skill-name>, <output-directory>, and skill purpose descriptions in SKILL.md workflows.
Boundary markers: None are implemented in the generated SKILL.md templates or the scripting logic.
Capability inventory: File system directory creation (mkdir), file writing (Path.write_text), and archive creation (zipfile.ZipFile) performed by the Python scripts and Write Bash tool.
Sanitization: The scripts/init_skill.py script includes a validate_skill_name function that restricts skill names to alphanumeric characters and hyphens (kebab-case). No sanitization is applied to broader 'purpose' descriptions.
[DATA_EXFILTRATION]: The scripts/package_skill.py script archives the contents of a specified directory. While it explicitly skips files starting with a dot (e.g., .env, .bashrc), it will include sensitive files located inside hidden directories (e.g., credentials within a .aws folder) if the agent is misled into packaging such locations.

skill-creator