fleet-race

Runs three-machine competitive analysis with cross-rebuttal and chief synthesis.

2|Updated Apr 5, 2026
One-click install
npx skills add https://github.com/supportersimulator/multi-fleet --skill fleet-race
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: fleet-race
Source: https://github.com/supportersimulator/multi-fleet/tree/main/skills/fleet-race
Command: npx skills add https://github.com/supportersimulator/multi-fleet --skill fleet-race

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Competitive branch racing across three machines enables independent exploration of solution approaches. It provides a structured protocol where each node works in isolated worktrees, submits a local_verdict, then engages in pairwise cross-rebuttal before the chief synthesizes results to converge on the strongest solution.

Core Features & Use Cases

  • Isolated worktrees per node for independent exploration and safe experimentation.
  • Pairwise cross-rebuttal and chief synthesis to surface disagreements and converge on a winner.
  • End-to-end governance including timeouts, evidence-based decisions, and optional escalation to human review.
  • Triggers for initiating a race (e.g., "race on this task", "fleet race") and integration with fleet tooling.

Quick Start

Configure three machines, launch the race task to each node, and monitor verdicts as they are submitted.

Frequently Asked Questions about fleet-race

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
What is fleet racing for architecture decisions and complex refactors?

Fleet racing is a distributed-systems protocol where three machines explore rival branches in isolated worktrees, submit local verdicts, and engage in pairwise cross-rebuttal to converge on the strongest solution.

How do I run a branch race across three machines?

Configure three machines with isolated worktrees, trigger the race task to each node, and monitor as local verdicts are submitted. The chief then synthesizes results through cross-rebuttal to automate a merge or escalate.

Can I use isolated worktrees to compare different performance profiling approaches?

Yes, you can apply fleet racing to performance profiling by assigning each node an isolated worktree for independent exploration, followed by cross-rebuttal and chief synthesis to determine the optimal profiling solution.

What is the best way to resolve disagreements between rival branches in git-ops?

The best way is pairwise cross-rebuttal followed by chief synthesis. Each node submits a local_verdict, disagreements are surfaced through rebuttal, and the chief aggregates verdicts to converge on the winning solution.

Does fleet racing support automated merge and human escalation for complex refactors?

Yes, fleet racing provides end-to-end governance including timeouts and evidence-based decisions. The chief synthesis automatically merges the winning branch or optionally escalates to human review if needed.

When should I not use cross-rebuttal branch synthesis for distributed systems?

You should avoid cross-rebuttal branch synthesis for simple tasks that do not require isolated exploration across three nodes, as the overhead of local verdicts and chief synthesis outweighs the benefit of converging on a solution.