The Agent Skills Directory

[SAFE]: The skill manages project tasks by performing read-only operations on task metadata files like task.json and task.md. It utilizes structured processing with jq to handle this data securely without exposing the system to injection or data exfiltration risks.
[COMMAND_EXECUTION]: Execution of bash commands is restricted through the allowed-tools configuration to a specific set of internal scripts (next-actions.sh and task-actions.sh). This restricted scope prevents the execution of arbitrary or unauthorized commands.
[PRIVILEGE_ESCALATION]: The skill uses the EnterWorktree tool to switch between development environments. This tool includes built-in safety mechanisms that require explicit user approval when accessing paths outside of the standard worktree directory.

next