Token导航 LogoToken导航

Agent Skill · 研究检索

gsdl-execute-plan(gsdl 执行计划)

gsdl-execute-plan 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。

gsdl-execute-plan 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,权限为接触敏感数据;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 45,663。

总安装 832周安装 34GitHub Stars 3下载 269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:gsdl-execute-plan(gsdl 执行计划)
来源仓库:https://github.com/nsantini/gsdl
仓库路径:skills/gsdl-execute-plan
安装命令:
npx skills add https://github.com/nsantini/gsdl --skill gsdl-execute-plan
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/nsantini/gsdl --skill gsdl-execute-plan

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。

gsdl-execute-plan 用于查找、检索和筛选相关信息。适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词、任务场景或来源线索快速定位候选结果时使用。通过 npx skills add 命令从指定 GitHub 仓库安装,需结合原始 README 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操作。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/gsdl-execute-plan

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/nsantini/gsdl --skill gsdl-execute-plan;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

    它需要接触敏感数据,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

gsdl-execute-plan 怎么安装?

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

gsdl-execute-plan 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。

gsdl-execute-plan 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

gsdl-execute-plan 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。

gsdl-execute-plan 用的人多吗?

公开数据是累计安装 832 次,近一周安装 34 次,下载 269 次,GitHub 3 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。

gsdl-execute-plan 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

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。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 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

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 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
gsdl-execute-plan

gsdl 执行计划 · 本页

8323接触敏感数据未明确通过

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

Manage Task List

Guides implementation of task lists with structured completion tracking. This skill enforces a disciplined, step-by-step approach to working through tasks generated from PRDs.

Project Context

Task lists are located at .planning/[project-name]/tasks-[prd-name].md. Implementation code lives at the workspace root (not inside .planning/).

Operating Modes

Standard Mode (default)

Work on ONE sub-task at a time. Do NOT start the next sub-task until the user explicitly approves. Stop after each sub-task and wait for "yes", "go", "next", etc.

Batch Mode (GSD orchestrator only)

If your instructions say "BATCH MODE", complete all sub-tasks under your assigned parent task without pausing for approval between sub-tasks. Apply the full completion protocol (mark [x], update task file) after each sub-task, then immediately continue to the next. When the entire parent task is [x], follow the Parent Task Completion protocol (including the git commit), then stop and return a summary — do not start any other parent task.

Completion Protocol

Follow this protocol strictly when completing tasks:

When You Finish a Sub-Task

  1. Mark the sub-task as completed: Change [] to [x] immediately
  2. Update the task list file: Save the changes to the task list
  3. Check parent task: If ALL subtasks under a parent are now [x], mark the parent as [x] too, then follow the Parent Task Completion protocol below
  4. Pause *(Standard Mode only)*: Stop and wait for user approval before starting the next sub-task. In Batch Mode, skip this step and continue immediately.

When You Complete a Parent Task

Once all sub-tasks under a parent are [x] and the parent itself is marked [x]:

  1. Stage all changes: git add -A
  2. Commit with a descriptive message:
git commit -m "$(cat <<'EOF'
[N.0] [Parent Task Title]

[2–4 bullet points summarising what was implemented, one per sub-task or logical group]
EOF
)"

Example:

git commit -m "$(cat <<'EOF'
1.0 Set up authentication module

- Created auth directory structure and base config
- Installed and configured JWT dependencies
- Added environment variable definitions for secrets
EOF
)"
  1. Then pause for user approval (Standard Mode) or continue to the next parent task (Batch Mode).

Example Workflow

Before:
- [ ] 1.0 Parent Task
  - [x] 1.1 First sub-task (completed earlier)
  - [ ] 1.2 Second sub-task (just finished)
  - [ ] 1.3 Third sub-task (not started)

After (when 1.2 is complete):
- [ ] 1.0 Parent Task
  - [x] 1.1 First sub-task
  - [x] 1.2 Second sub-task
  - [ ] 1.3 Third sub-task (next up)

After (when 1.3 is complete and parent is done):
- [x] 1.0 Parent Task
  - [x] 1.1 First sub-task
  - [x] 1.2 Second sub-task
  - [x] 1.3 Third sub-task

Task List Maintenance

Update as You Work

  1. Mark completed items: Update [x] for each finished task/sub-task
  2. Add new tasks: If you discover additional work needed, add new tasks to the list
  3. Keep files current: Maintain the "Relevant Files" section with accurate descriptions

Relevant Files Section

The "Relevant Files" section should be kept up to date:

  • List every file created or modified (implementation files at workspace root, planning files under .planning/)
  • Use full paths relative to workspace root (e.g., src/file.ts for code, .planning/[project-name]/tasks-prd-name.md for planning)
  • Provide a one-line description of each file's purpose
  • Add new files as they are created during implementation

Implementation Process

Before Starting Work

  1. Read the task list from disk: Read the task file directly from its path on disk — do not rely on any in-context version. The user may have edited the file (added tasks, reworded sub-tasks, reordered items) since the plan was generated. Always use the on-disk state as the source of truth.
  2. Identify the next sub-task: Find the first unchecked [] sub-task
  3. Check for dependencies: Ensure previous tasks are completed
  4. Understand the goal: Make sure you understand what the sub-task requires

During Implementation

  1. Focus on current sub-task: Work only on the current sub-task
  2. Implement thoroughly: Write code, tests, and documentation as needed
  3. Test your work: Verify the implementation works correctly

After Completing a Sub-Task

  1. Update task list: Mark the sub-task as [x]
  2. Check parent task: If all sub-tasks done, mark parent as [x], then follow the Parent Task Completion protocol (commit before proceeding)
  3. Update Relevant Files: Add any new files created
  4. Save the task list file: Persist the changes
  5. Report to user: Briefly describe what was completed
  6. Request permission: Ask "Ready to move to the next sub-task?" or similar
  7. Wait: Do not proceed until the user confirms

AI Instructions

When working with task lists, you must:

  1. Regularly update the task list file after finishing any significant work
  2. Follow the completion protocol:

- Mark each finished sub-task [x] - Mark parent task [x] once all its subtasks are [x]

  1. Add newly discovered tasks if additional work is identified
  2. Keep "Relevant Files" accurate and up to date
  3. Check which sub-task is next before starting work
  4. Pause after each sub-task and wait for user approval

User Permission Phrases

Accept these as permission to continue to the next sub-task:

  • "yes"
  • "y"
  • "go"
  • "continue"
  • "next"
  • "proceed"
  • "keep going"

Do NOT continue if the user:

  • Asks questions
  • Requests changes
  • Wants to review something
  • Says "wait", "hold on", "stop", etc.

Handling Changes and Additions

Discovered Issues

If you discover issues while implementing:

  1. Add a new task or sub-task to the list
  2. Inform the user about the addition
  3. Ask if they want to address it now or later

Task Modifications

If the user requests changes to the task list:

  1. Update the task list as requested
  2. Confirm the changes with the user
  3. Resume from the current position

Skipping Tasks

If the user wants to skip a task:

  1. Mark it with a note (e.g., - [] 2.3 [SKIPPED] Original description)
  2. Continue to the next task
  3. Update the parent task status appropriately

Quality Focus

This one-task-at-a-time approach ensures:

  • Thoroughness: Each task gets full attention
  • Quality: Work is reviewed before moving on
  • Alignment: User can course-correct at each step
  • Documentation: Progress is clearly tracked
  • Learning: Junior developers see proper implementation pace

Example Interaction

AI: "I've completed sub-task 1.2: Create login endpoint. The endpoint is now functional with proper error handling. I've marked it as complete in the task list. Ready to move to sub-task 1.3: Implement credential validation?"

User: "yes"

AI: [Proceeds to work on sub-task 1.3]

File Location

Task lists follow this structure:

  • Path: .planning/[project-name]/tasks-[prd-name].md
  • Example: .planning/my-auth-project/tasks-prd-user-authentication.md

Always work within the project's directory structure.

平台分布

Codex

36.83%
按下载量换算99

Claude

27.52%
按下载量换算74

Cursor

19.04%
按下载量换算51

Gemini CLI

8.87%
按下载量换算24

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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。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