Generate parallel candidate implementations in isolated worktrees and apply the best result.
npx skills add https://github.com/garochee33/DSH --skill best-of-n-garochee33
Please help me install this Agent Skill. Skill: best-of-n Source: https://github.com/garochee33/DSH/tree/main/agents/grok/skills/best-of-n Command: npx skills add https://github.com/garochee33/DSH --skill best-of-n-garochee33
This Skill helps you solve a task by exploring multiple implementations in parallel, so you can choose the strongest result instead of settling for the first idea.
Ask the skill to solve your task with three parallel candidates and return the best implementation.
High-intent search queries and answers about installing and using this skill.
To compare refactoring approaches, generate multiple parallel implementations in isolated worktrees, evaluate them for correctness and safety, then merge the winning candidate back into your main workspace.
The best way to handle uncertain design choices is spinning up independent subagents to tackle the same problem in parallel, comparatively evaluating the results, and applying the strongest implementation.
Use parallel subagents by having each execute an independent fix in an isolated worktree, evaluating all candidates for code quality and correctness, and selecting the winning solution for application.
Yes, isolated worktrees are required for parallel candidate generation, ensuring each subagent can independently implement and evaluate its approach without interfering with the main workspace.
Avoid using a best-of-n approach for feature delivery when tasks are straightforward and lack uncertain design choices, because the overhead of spinning up parallel subagents and isolated worktrees outweighs the comparative evaluation benefit.
N0K0Implement GitHub issue checklists using isolated git worktrees and synced progress.
doravidanMerge RALPH worktree changes into the main branch with optional squash.
CaptainCrouton89Evaluate code changes for security, correctness, performance, and spec alignment.
tmalcolm-0607Orchestrate wave-based parallel feature implementation across isolated git worktrees.
