Back to Details

supervision

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Supervision

监督策略

Use this skill when the user wants to configure how much autonomy the agent should use for the current project.
当用户想要配置当前项目中Agent应拥有的自主权限程度时，使用此技能。

Scope

适用范围

This skill manages 
## Supervision Policy
in 
RESEARCH.md
.
If the user wants to configure both supervision and preferred skillpacks together, prefer 
customize
and keep this skill focused on supervision only.
本技能负责管理
RESEARCH.md
中的
## Supervision Policy
章节。
如果用户希望同时配置监督策略和偏好的skillpacks，请优先使用
customize
技能，本技能仅专注于监督策略配置。

Flow

流程

Use a preset-first flow:
  1. choose a preset: 
    manual
    , 
    checkpointed
    , 
    autonomous
    , or 
    wild
  2. confirm notification events
  3. confirm approval gates
  4. confirm stop target / limits
  5. confirm resource rules
  6. confirm idea-change handling
  7. write the policy into 
    RESEARCH.md
采用预设优先流程：
  1. 选择一个预设模式：
    manual
    、
    checkpointed
    、
    autonomous
    或
    wild
  2. 确认通知事件
  3. 确认审批关卡
  4. 确认停止目标/限制条件
  5. 确认资源规则
  6. 确认想法变更处理方式
  7. 将策略写入
    RESEARCH.md

Precedence

优先级

RESEARCH.md
## Supervision Policy
is the live working-state layer and overrides everything. 
.co-researcher/skills.yaml
supervision
is the project-default layer written by 
customize
— only consulted when 
RESEARCH.md
has no policy yet.
RESEARCH.md
中的
## Supervision Policy
是实时生效的工作状态层，优先级高于所有其他配置。 
.co-researcher/skills.yaml
中的
supervision
是由
customize
技能写入的项目默认层——仅当
RESEARCH.md
中尚未存在策略时才会被参考。

Constraints

约束条件

  • Do not create a separate hidden config store for supervision.
  • Keep the policy human-readable.
  • If 
    RESEARCH.md
    is missing, initialize it before editing.
  • wild
    is never implicit; require explicit user confirmation.
  • 请勿为监督策略创建单独的隐藏配置存储。
  • 确保策略具备人类可读性。
  • 如果
    RESEARCH.md
    不存在，请先初始化再进行编辑。
  • wild
    模式绝不会被默认启用；必须获得用户的明确确认。