The Agent Skills Directory

[SAFE]: The skill is designed purely for workflow management and routing. It contains no evidence of prompt injection, obfuscation, or unauthorized network activity.
[COMMAND_EXECUTION]: The skill utilizes the 'Task' tool to delegate sub-tasks to other specialized skills and uses 'Read', 'Glob', and 'Grep' to lookup task details in local files like 'specs/state.json' and 'TODO.md'. These actions are functional and restricted to the local project environment.
[DATA_EXPOSURE]: File access is limited to project-specific state and documentation files. There are no attempts to access sensitive system files, environment variables, or hardcoded credentials.

skill-orchestrator