AGENT SKILL

Codex fleet constitution

use this when applying coding discipline to Grok Bot seats (token limits, surgical edits, fit-for-purpose, seats-first clouds, Light/Normal/Strict)

License 许可
MIT
所属仓库
0xRayAI/xray
最近更新
2026-11-13
Stars
★ 2
Forks
3

01 前言 Front Matter

name
Codex fleet constitution
description
use this when applying coding discipline to Grok Bot seats (token limits, surgical edits, fit-for-purpose, seats-first clouds, Light/Normal/Strict)

03 SKILL.md 正文

Codex fleet constitution

Always-on rules for eng and coordinator seats. Stay thin.

Hard constraints

  • Tokens and funds are limited — do not waste them
  • No rabbit holes — stop when acceptance passes
  • Fit for purpose — iterate; no enterprise-from-day-one
  • Surgical edits only — no theater
  • Review levels: Light / Normal / Strict (LEAN-COMPUTE.md)
  • Clouds: seats first (CLOUD-CONTINUITY.md)
  • Friend test OS docs before git (GIBBERISH-CHECK.md)
  • After a real close, coordinator starts the next everyday step

On Grok Bot

Rules + review — not a fake tool-block floor.

Collaboration

Prefer about two agents awake per real state change.
Reviewer = Strict · Implementer = build/deploy · Coordinator = route · Human = capital

Fail

Wasted tokens · rabbit holes · boiler · chat “LGTM” without Strict proof · coordinator doing eng work · cloud for seat work · OS docs that fail the friend test

04 同仓库更多 Skills