Agent Skills › xai-org/grok-build › best-of-n

best-of-n

GitHub

并行生成N种实现方案并择优应用。解析任务与候选数，隔离spawn子代理，等待完成后按正确性、代码质量和安全性评估，展示对比表后应用最优解。

crates/codegen/xai-grok-shell/skills/best-of-n/SKILL.md xai-org/grok-build

触发场景

best of n try multiple approaches parallel implementations /best-of-n /bon

安装

npx skills add xai-org/grok-build --skill best-of-n -g -y

更多选项

非标准路径

npx skills add https://github.com/xai-org/grok-build/tree/main/crates/codegen/xai-grok-shell/skills/best-of-n -g -y

不安装直接使用

npx skills use xai-org/grok-build@best-of-n

指定 Agent (Claude Code)

npx skills add xai-org/grok-build --skill best-of-n -a claude-code -g -y

安装 repo 全部 skill

npx skills add xai-org/grok-build --all -g -y

预览 repo 内 skill

npx skills add xai-org/grok-build --list

SKILL.md

Frontmatter

{
    "name": "best-of-n",
    "metadata": {
        "short-description": "Parallel implementation tournament"
    },
    "description": "Implement a task N ways in parallel and pick the best. Spawns multiple subagents in isolated worktrees, evaluates all candidates, and applies the winner. Use when asked to \"best of n\", \"try multiple approaches\", \"parallel implementations\", \"\/best-of-n\", or \"\/bon\".\n"
}

/best-of-n -- Parallel Implementation Tournament

Implement a task multiple different ways in parallel, evaluate all candidates, and apply the best one.

Usage

/best-of-n [N] <task>

If the first token is a number 2-10, it sets the candidate count; the rest is the task.
If omitted, N defaults to 3.

Examples:

/best-of-n implement the login page (3 candidates)
/best-of-n 5 refactor the auth module (5 candidates)

Steps

Parse the user's message to extract N (candidate count, default 3) and the task description.
Spawn N subagents in a single message (parallel tool calls). Use the task tool for each with:
- subagent_type: "general-purpose"
- isolation: "worktree"
- run_in_background: true
- description: "Candidate <number>"
- prompt: the task description, plus "You are candidate <number> of <N> independent implementations. Implement the task fully. When done, summarize your approach and the changes you made."
Wait for all candidates to complete using get_task_output with block: true or wait_tasks with mode: "wait_all".
Evaluate and pick the winner using the criteria below.
Apply the winner's changes from its worktree to the main workspace. Review the changes in context and fix any remaining issues.
End your response with WINNER: <number> (1-N).

Evaluation Criteria

Evaluate each candidate on these axes, in order of importance:

Correctness -- Does the candidate actually solve the task? Does it handle the requirements completely, or does it miss important aspects? Are there logic errors, type errors, or broken imports?
Code Quality -- Is the code clean, readable, and well-structured? Does it follow the patterns and conventions of the surrounding codebase? Does it avoid unnecessary complexity?
Safety -- Does the candidate avoid introducing bugs, security issues, or breaking changes to existing functionality?

How to Decide

Focus on correctness first. A candidate that fully solves the task with minor style issues beats one that is beautifully written but incomplete or wrong.
If multiple candidates are equally correct, prefer the one with cleaner code and better codebase integration.
If a candidate introduces unnecessary changes beyond the task scope, count that against it.
If all candidates are poor, still pick the least bad one.