The Agent Skills Directory

[SAFE]: The skill follows established patterns for artifact creation within the framework. All file operations are restricted to the local project directory, specifically the .claude/ folder and skill-specific directories.
[SAFE]: State management hooks (pre-execute.cjs, post-execute.cjs) properly track the lifecycle of the creator by writing to a local runtime state file (active-creators.json). This is a legitimate internal coordination mechanism.
[SAFE]: The primary script (main.cjs) implements safe JSON parsing and type inference logic. It avoids dynamic execution functions such as eval() or exec() and relies solely on standard Node.js fs and path modules for local file management.
[SAFE]: Research requirements mentioned in the documentation target well-known academic and technology services (e.g., arXiv), which are considered trusted sources according to security guidelines.

schema-creator