Token导航 LogoToken导航

Agent Skill · 研究检索

ghpm-init(ghpm 初始化)

ghpm-init 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。

ghpm-init 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 47,712。

总安装 722周安装 31GitHub Stars 19下载 253核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

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:ghpm-init(ghpm 初始化)
来源仓库:https://github.com/jackchuka/ghpm
仓库路径:skills/ghpm-init
安装命令:
npx skills add https://github.com/jackchuka/ghpm --skill ghpm-init
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/jackchuka/ghpm --skill ghpm-init

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

GHPM 初始化技能用于信息查找、检索和筛选,支持关键词快速定位。适用于需要根据任务场景或来源线索定位候选结果的场景。通过 GitHub 仓库安装,使用 npx skills add 命令添加技能模块。安装前建议确认权限范围和维护状态,注意可能触发的网络请求。

ghpm-init 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/ghpm-init

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/jackchuka/ghpm --skill ghpm-init;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

ghpm-init 怎么安装?

在终端执行 npx skills add https://github.com/jackchuka/ghpm --skill ghpm-init。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

ghpm-init 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

ghpm-init 需要什么权限?

权限级别是需要联网。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。

ghpm-init 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

ghpm-init 用的人多吗?

公开数据是累计安装 722 次,近一周安装 31 次,下载 253 次,GitHub 19 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

ghpm-init 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
ghpm-init

ghpm 初始化 · 本页

72219需要联网未明确有提醒

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

ghpm-init

PREREQUISITE: Read ../ghpm-shared/SKILL.md for prerequisites and error handling.

Initialize a GitHub Projects v2 project for use with ghpm skills. Auto-discovers schema and generates config files.

Arguments

  • Project URL (e.g., https://github.com/orgs/<org>/projects/<number>)
  • If no arguments, ask the user for the project URL or owner + number.

Workflow

Phase 1: Validate Access

  1. Parse arguments to extract owner and project number from URL or positional args.
  2. Run gh auth status to verify authentication. If this fails, tell the user to run gh auth login and stop.
  3. Run gh project view <number> --owner <owner> --format json to verify access and get project metadata (id, title). If this fails with 403/404, tell the user to check token scopes (read:project, project) and project access.
  4. Determine the owner type from the project view output. This matters for Phase 3: use organization(login: "<owner>") for org-owned projects or user(login: "<owner>") for user-owned projects.

Phase 2: Discover Schema

  1. Run gh project field-list <number> --owner <owner> --format json to get all fields with IDs, types, and options.
  2. Identify the workflow field: find the ProjectV2SingleSelectField named "Status". If multiple candidates or none found, ask the user which field drives the kanban workflow.
  3. Separate the workflow field from other fields.

Phase 3: Discover Views

  1. Run GraphQL query to fetch all views. Use the owner type determined in Phase 1 to choose the correct root field:
gh api graphql -f query='
{
  organization(login: "<owner>") {
    projectV2(number: <number>) {
      views(first: 50) {
        nodes {
          id
          name
          number
          layout
          filter
          groupByFields(first: 5) {
            nodes {
              ... on ProjectV2Field { name }
              ... on ProjectV2SingleSelectField { name }
              ... on ProjectV2IterationField { name }
            }
          }
          sortByFields(first: 5) {
            nodes {
              field {
                ... on ProjectV2Field { name }
                ... on ProjectV2SingleSelectField { name }
                ... on ProjectV2IterationField { name }
              }
              direction
            }
          }
        }
      }
    }
  }
}'
  1. Transform view data into config format: lowercase layout names (TABLE_LAYOUT -> table, BOARD_LAYOUT -> board, ROADMAP_LAYOUT -> roadmap), preserve filter strings as-is.

Phase 4: Discover Items and Repos

  1. Run gh project item-list <number> --owner <owner> --format json --limit 1000 to fetch all items. The --limit flag controls the total count returned (no cursor pagination available), so use a high value. Note: items beyond the limit are silently truncated — see cache reference for details.
  2. Extract unique repository URLs from items to build the repos list.
  3. Store all items in cache.

Phase 5: Write Config Files

  1. Assemble .ghpm/config.json with this structure:
{
  "version": 1,
  "project": {
    "owner": "<owner>",
    "number": "<number>",
    "id": "<project_node_id>",
    "title": "<project_title>"
  },
  "workflow": {
    "field": "<status_field_name>",
    "field_id": "<status_field_id>",
    "columns": [{ "id": "<option_id>", "name": "<option_name>" }]
  },
  "fields": [
    {
      "name": "<field_name>",
      "id": "<field_id>",
      "type": "<field_type>",
      "options": [{ "id": "<option_id>", "name": "<option_name>" }]
    }
  ],
  "views": [
    {
      "name": "<view_name>",
      "number": "<view_number>",
      "id": "<view_id>",
      "layout": "<table|board|roadmap>",
      "filter": "<filter_string>",
      "sort": [{ "field": "<field_name>", "direction": "<ASC|DESC>" }],
      "group_by": ["<field_name>"]
    }
  ],
  "cache": {
    "ttl_minutes": 30
  },
  "conventions": {
    "branch": "{type}/{issue-number}/{slug}. Detect type from issue labels and title: bug→fix, enhancement/feature→feat, documentation→docs, test→test, refactor→refactor, default→chore",
    "status_sync": "On start, set to InProgress if currently Planned or ReadyForDev. On session end with merged PR, set to Done.",
    "decisions": "When a design decision is detected, nudge before recording. Post as issue comment."
  },
  "repos": ["<owner>/<repo>"]
}
  • Only include sort key on views that have sort fields.
  • Only include group_by key on views that have groupBy fields.
  • Only include options on fields that are single_select or iteration type.
  • Strip the full URL from repos, keep just owner/repo.
  1. Write .ghpm/cache.json per ../ghpm-shared/references/cache.md.
  2. Run mkdir -p.ghpm/sessions. Check if .gitignore exists. If it does, check if .ghpm/ is already listed. If not, ensure the file ends with a newline before appending .ghpm/ on its own line. If .gitignore doesn't exist, create it with .ghpm/\n.

Phase 6: Install Agent Integration (optional)

  1. Detect the current agent environment and offer to install hooks/rules for enhanced ghpm-work sessions. See ../ghpm-shared/references/integrations.md for available integrations. Claude Code detection: Check if .claude/ directory exists or if the current process is running inside Claude Code. If detected, ask the user: Detected Claude Code. Install session hooks for ghpm-work? (y/n) Hooks add: auto session context injection, stale session detection. Without hooks, ghpm-work still works via explicit commands. If confirmed: If declined or agent not detected, skip.

- Add the hooks from ../ghpm-shared/references/integrations.md into .claude/settings.local.json under the "hooks" key. - If .claude/settings.local.json already exists, merge the "hooks" key into the existing structure (preserve existing settings and hooks, add new ones).

Phase 7: Report

  1. Print summary:
ghpm initialized for <title> (<owner>/projects/<number>)

  Fields:    <count> (<list of field names>)
  Views:     <count> (<list of view names>)
  Repos:     <count>
  Items:     <count> cached

Config:  .ghpm/config.json (gitignored)
Cache:   .ghpm/cache.json (gitignored)
Hooks:   <installed|skipped>

Run /ghpm-status for project overview.

See Also

平台分布

Codex

34.97%
按下载量换算88

Claude

28.58%
按下载量换算72

Cursor

18.35%
按下载量换算46

Gemini CLI

8.61%
按下载量换算22

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills