back
loading skill details...
Dispatch parallel agents.
Spawn isolated agents for parallel task execution with runtime-native team selection and fresh context per wave. Auto-selects spawn backend based on runtime: Claude Native Teams in Claude sessions, Codex sub-agents in Codex sessions, or inline fallback if multi-agent unavailable Executes tasks in dependency-ordered waves, spawning only unblocked tasks in parallel while respecting file ownership conflicts via pre-spawn manifest validation Implements Ralph Wiggum pattern: fresh worker context per wave, TaskList as persistent state, filesystem for all artifacts, messaging for coordination only Supports worktree isolation for multi-epic dispatch, OL wave integration with completion backflow, and per-task or per-wave commit policies Includes scope-escape protocol for workers discovering out-of-bounds work and merge arbiter for conflict resolution during worktree reintegration Swarm Skill Spawn isolated agents to execute tasks in parallel. Fresh context per agent (Ralph Wiggum pattern). Loop position Move 5 (wave execution) of the operating loop, specifically the parallel-fork primitive /crank invokes. Refuses to spawn parallel agents on a wave that has not cleared the wave-validity check in the slice validation plan: write scopes must be disjoint, no shared migration/contract/CLI surface, integration order declared when it matters, one owner per slice, discard path per slice. Parallelism is explicit ownership, not swarm chaos. Default to sequential when the wave-validity rows are not all green. Integration modes: Direct - Create TaskList tasks, invoke /swarm Via Crank - /crank creates tasks from beads, invokes /swarm for each wave Requires multi-agent runtime. Swarm needs a runtime that can spawn parallel subagents. If unavailable, work must be done sequentially in the current session. Architecture (Mayor-First)
don't have the plugin yet? install it then click "run inline in claude" again.