The Agent Skills Directory

[SAFE]: The skill implements procedural scripts to manage the lifecycle of agent skills, performing only legitimate file system operations consistent with its stated purpose.\n- [SAFE]: The init_skill.py script sanitizes the skill name input via the normalize_skill_name function, preventing potential directory traversal or template injection issues from user-provided input.\n- [SAFE]: The package_skill.py utility provides security safeguards by explicitly rejecting symbolic links and verifying that files are contained within the intended skill root before they are added to the packaged archive.\n- [SAFE]: The validation logic in quick_validate.py defaults to safe parsing practices for metadata extraction by using yaml.safe_load when PyYAML is available, mitigating risks associated with unsafe deserialization.\n- [SAFE]: Regarding potential indirect prompt injection surface (Category 8): (1) Ingestion points: skill_name and output path in init_skill.py; (2) Boundary markers: N/A for local CLI-style automation; (3) Capability inventory: scripts/init_skill.py and scripts/package_skill.py perform file writing and directory creation; (4) Sanitization: The normalize_skill_name function removes non-alphanumeric characters. No exploitable vulnerabilities were identified in this surface.

skill-creator