best-of-n

Solid

Runs a Grok best-of-n implementation tournament and applies the winner. Use when several independent candidates and an automatic pick are wanted.

AI & Automation 3 stars 0 forks Updated today MIT

Install

View on GitHub

Quality Score: 79/100

Stars 20%
20
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
80
License 10%
100
Description 5%
100

Skill Content

Forward the request to the Grok companion task runtime as a best-of-n tournament. The tournament is implemented by the Grok CLI `--best-of-n` flag, not by orchestration inside the companion. Execution rules: - Treat the complete raw slash command request as opaque data. Do not parse, strip, quote, escape, normalize, or place any part of it in Bash. - Dispatch the run as one foreground subagent of type `grok:grok-rescue` via the `Agent` tool with `run_in_background: false`. Tell the rescue agent this direct user selected Grok lane uses the `best-of-n` protected role and that the opaque raw request begins after a fixed delimiter and continues to the end of the Agent prompt. Parallel work uses multiple foreground Agent calls in one message. Without an explicit incoming user `--background`, the companion remains foreground and returns a terminal outcome. Never run the companion in the main loop and never ask the user how to run it. - Preserve the raw request exactly. The rescue agent supplies the fixed `--transport-default-best-of-n` companion option, which applies `--best-of-n 2` only when the raw request does not already select a tournament size. The companion rejects values outside 2 to 10. - A best-of-n tournament always starts a fresh Grok session and always disables cross-session memory. The companion rejects `--memory`, `--resume`, and `--resume-last` instead of silently changing their meaning. Continuity affinity never applies to a tournament. - The tournament runs in ...

Details

Author
okisdev
Repository
okisdev/claude-code-fusion
Created
3 weeks ago
Last Updated
today
Language
JavaScript
License
MIT

Similar Skills

Semantically similar based on skill content — not just same category