goal-loop

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Goal Loop

目标循环

Steps

步骤

  1. Parse goal. Read
    max_rounds
    (default
    5
    ). Reject non-positive integer.
  2. Infer completion criteria from goal and repo context. Record uncertainty as blocker only when it prevents safe next action.
  3. Choose smallest useful next action. Route through existing skill when one fits.
  4. Verify with narrowest useful check: test, diff, command output, artifact.
  5. Critique: what proves or disproves completion? What remains? Smallest next action?
  6. Criteria evidenced: stop
    completed
    . max_rounds reached: stop
    max_rounds
    . Else repeat from 3.
  1. 解析目标。读取
    max_rounds
    (默认值为
    5
    )。拒绝非正整数。
  2. 根据目标和 repo 上下文推断完成标准。仅当不确定性阻碍安全的下一步操作时,才将其记录为 blocker。
  3. 选择最小且有用的下一步操作。如果有合适的现有 skill,则通过该 skill 执行。
  4. 使用范围最小且有用的检查进行验证:test、diff、命令输出、artifact。
  5. 批判性审视:什么能够证明或推翻完成状态?还剩下什么?下一步最小操作是什么?
  6. 完成标准已有证据:停止并返回
    completed
    。达到
    max_rounds
    :停止并返回
    max_rounds
    。否则从第 3 步重复。

Rules

规则

A round = one state-changing action + verification. Tool calls are not rounds. Do not spend rounds restating goal, polling without change, or speculative cleanup. Pause before commits, pushes, PR, deploy, production, credentials, irreversible ops.
一轮(round)= 一次改变状态的操作 + 一次验证。工具调用不算轮次。不要浪费轮次复述目标、无变化地轮询或进行推测性清理。在 commit、push、PR、deploy、production、credentials 或不可逆操作之前暂停。

Status

状态

End with exactly one:
completed
,
blocked
(name unblocker),
max_rounds
(name remaining criteria).
最后必须且只能返回以下之一:
completed
、
blocked
(注明解除阻塞所需的条件)、
max_rounds
(注明尚未满足的标准)。