Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计提醒

upgrade-chief升级酋长

Agent Skill

upgrade-chief 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

384

周安装

16

GitHub Stars

26

下载量

128
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:upgrade-chief(升级酋长)
来源仓库:https://github.com/thaitype/chief-agent-framework
仓库路径:skills/upgrade-chief
安装命令:
npx skills add https://github.com/thaitype/chief-agent-framework --skill upgrade-chief
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/thaitype/chief-agent-framework --skill upgrade-chief

简介

用于查找、检索和筛选相关信息。upgrade-chief 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

  • 适合在需要根据关键词或任务场景快速定位候选结果时使用。
  • 可结合来源仓库、安装命令和原始 README 继续核验具体用法。
  • 安装前建议确认权限范围、维护状态,以及是否会触发联网或文件读写。
  • 注意该技能当前分类为研究检索,功能聚焦架构或系统升级。

SKILL.md

Upgrade the Chief Agent Framework to the version specified in the arguments.

Arguments

The first argument is the target version (branch or tag). Optional.

  • No argument → upgrade to the latest stable release (highest semver tag). Find it by running git ls-remote --tags https://github.com/thaitype/chief-agent-framework.git, strip refs/tags/, ignore ^{} entries, and pick the highest semver version.
  • canary → latest canary branch (active development, unreleased)
  • v1.0.0, v2.0.0, etc. → specific tagged version

Steps

0. Detect coding agent and install mode

Before cloning, detect which coding agent the user has set up by checking directories:

  1. Check if .claude/agents/ exists → suggest Claude Code
  2. Check if .github/agents/ exists with agent files → suggest Copilot
  3. If only .agents/ exists (no coding-agent-specific directory) → suggest OpenCode (or unknown agent)

For coding agents that use their own directory, detect install mode:

  1. Check if any file in the coding-agent-specific directory is a symlink → current mode is link
  2. If files are regular files → current mode is copy
  3. If no coding-agent-specific directory exists → no mode detected

Always ask the user to confirm:

  • Which coding agent they use — Supported agents: claude-code, opencode, codex, cursor, copilot, gemini-cli, amp, windsurf, kiro, aider
  • Which install mode they want: link (recommended) or copy

Default to the detected coding agent and mode. If no mode is detected, suggest link. On Windows, if symlinks are not available (Developer Mode disabled), suggest copy mode. OpenCode reads .agents/ directly and needs no install mode.

1. Clone the target version

git clone --depth 1 --branch <version> https://github.com/thaitype/chief-agent-framework.git .chief-agent-tmp

2. Diff current files vs target version

Compare these paths between the current project and .chief-agent-tmp/template/:

Note: The framework repo stores installable files under template/. When comparing, map .chief-agent-tmp/template/<path> to <path> in the current project.

  • AGENTS.md (or CLAUDE.md if AGENTS.md does not exist)
  • .agents/agents/*.md
  • .agents/skills/*/
  • .chief/MANUAL.md
  • .chief/project.md (template only — compare structure, not user content)
  • .chief/_rules/ (directory structure, not user content)
  • .chief/_template/

3. Categorize each change

For every file that differs, classify it:

CategoryDescription
Framework updateFile exists in both, only target version changed. Safe to overwrite.
New fileFile exists in target but not in current project. Safe to add.
User-modified conflictFile exists in both, and the user has modified it from the original. Needs review.
Structure changeFile type changed (e.g. real file became symlink, directory renamed). Needs review.
RemovalFile exists in current but not in target. Check if user depends on it.

To detect user modifications: compare the current file against both the original installed version (if detectable from git history) and the target version. If uncertain, classify as conflict.

4. Present the upgrade plan

Format the plan clearly:

## Upgrade Plan: <current> → <target>

### Framework updates (safe to overwrite)
- <file> — <what changed>

### New files
- <file> — <description>

### Conflicts (needs review)
- <file> — you modified this file, target version also changed

### Structure changes
- <file> — <description of change>

### Removals
- <file> — exists locally but not in target version

If a section has no entries, omit it.

5. Wait for user approval

Ask the user to review the plan. Do NOT apply any changes until the user explicitly approves.

The user may:

  • Approve all changes
  • Approve some and reject others
  • Ask for more details about a specific change
  • Cancel the upgrade

6. Apply approved changes

For each approved change:

  • Framework update: overwrite the file from target version
  • New file: copy from target version
  • Conflict: show a diff and let the user decide (overwrite, skip, or merge manually)
  • Structure change: apply the structural change (e.g. create symlinks)
  • Removal: delete the file (only if user approves)

7. Update coding agent integration

After applying changes, check if new agents or skills were added in .agents/. Based on the coding agent detected in step 0:

Claude Code — update .claude/ using the chosen mode:

Link mode:

ln -s ../../.agents/agents/<new-agent>.md .claude/agents/<new-agent>.md
ln -s ../../.agents/skills/<new-skill> .claude/skills/<new-skill>

Copy mode:

cp .agents/agents/<new-agent>.md .claude/agents/<new-agent>.md
cp -r .agents/skills/<new-skill> .claude/skills/<new-skill>

Copilot — update .github/agents/ using the chosen mode:

Link mode:

ln -s ../../.agents/agents/<new-agent>.md .github/agents/<new-agent>.md

Copy mode:

cp .agents/agents/<new-agent>.md .github/agents/<new-agent>.md

Preserve the user's custom model values in existing agent files. Only update content, not the model: field, unless the user explicitly approves.

OpenCode — no action needed, it reads .agents/ directly.

Skip entries that already exist.

7b. Check model configuration (non-Claude Code only)

For non-Claude Code agents, check if agent files still contain the default model: opus or model: sonnet values (i.e., models were never customized by the user).

Determine the agent directory to check:

  • Copilot: check .github/agents/
  • Other non-Claude Code agents: check .agents/agents/

If any agent file still has placeholder model values (model: ${thinking_model} or model: ${coding_model}):

  • For Claude Code: auto-replace ${thinking_model}opus, ${coding_model}sonnet (no prompt needed)
  • For other agents: ask the user: Replace ${thinking_model} with the Thinking Model and ${coding_model} with the Coding Model.

1. Thinking Model (for chief-agent, e.g. o3, gemini-2.5-pro) 2. Coding Model (for builder/tester/review-plan, e.g. gpt-4.1, gemini-2.5-flash)

If all agent files already have real model values (no placeholders), skip this step — the user has already configured their models.

Also handle the coding-agent-specific rules file at root using the chosen mode:

  • For Claude Code: CLAUDE.md should be a symlink to AGENTS.md (link mode) or a copy (copy mode)
  • For Copilot: .github/agents/*.md should be symlinks (link mode) or copies (copy mode) from .agents/agents/*.md
  • If the current state doesn't match the chosen mode, include this in the upgrade plan as a structure change

8. Clean up

rm -rf .chief-agent-tmp

9. Summary

Report what was changed, what was skipped, and any manual steps the user still needs to take.

Important rules

  • NEVER apply changes without user approval
  • NEVER overwrite user content in .chief/ milestones (goals, contracts, plans, reports)
  • NEVER delete user files without explicit approval
  • If uncertain whether a file was user-modified, classify it as a conflict
  • Always clean up .chief-agent-tmp even if the upgrade is cancelled

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

40.02%
按下载量换算51

Claude

27.52%
按下载量换算35

Cursor

17.75%
按下载量换算23

Gemini CLI

9.7%
按下载量换算12

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills