Spawn parallel agents to solve problems

Spawn parallel subagents to compete on code, content, or research solutions, then merge the winner

When a problem benefits from diverse solution approaches and time permits parallel exploration with an objective win criterion.

• · problem: code optimization, content draft, research approach, or similar
• · N competing strategies (user-provided or agent-generated)

• · N parallel subagent solutions
• · evaluation by metric or LLM judge
• · merged winner into working tree

• · git DAG orchestration
• · message board coordination
• · stdlib Python tools
• · 7 slash commands
• · agent templates

• · git repo initialized + clean working tree
• · evaluation metric or judge criteria defined
• · problem statement clear enough to parallelize

• · N subagents diverge on problem interpretation, merging fails
• · Evaluation metric rewards local optimization over system quality
• · git DAG merge produces conflicts

• · 7 slash commands (spawn, evaluate, merge, compare, etc.)
• · git DAG orchestration (clean merge strategy)
• · message board coordination (visibility across agents)
• · agent templates (common patterns pre-built)