What problem does it solve? Coordinating multiple coding agents across sessions, tabs, and worktrees is error-prone: agents get spawned in the wrong session, prompts land on blocked panes, and long answers vanish inside alternate-screen TUIs. This Skill defines a policy for dispatching, prompting, monitoring, and sweeping agents through the Herdr CLI so delegated work stays organized and auditable. ## Core Features & Use Cases - Session pinning and targeting: Detect inside/outside host mode, pin a single Herdr session, and target agents by name or pane_id without hijacking the user's focused pane. - Agent lifecycle management: Start agents with model and permission flag presets, reuse live agents before spawning new ones, and send self-contained prompts with goal, location, constraints, and deliverable fields. - Monitoring and sweeps: Read agent state safely, surface blocked approval dialogs to the user, and run read-only sweeps grouped by session and status. - Use Case: You ask to delegate a bug fix on a new branch. The Skill pins the running session, creates a worktree and tab, starts a Cursor agent with the right model flags, waits for completion, and summarizes the Markdown report file. ## Quick Start Ask the assistant to spin up a coding agent on a specific task in the current repository and report back when it finishes.
