Token导航 LogoToken导航TokenDH.com
待分类需要联网github未标认证来源可访问许可证需确认审计通过

skill-create-workflow技能 create 工作流

Agent Skill

skill-create-workflow 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。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

196

周安装

8

GitHub Stars

公开资料未说明

下载量

63
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:skill-create-workflow(技能 create 工作流)
来源仓库:https://github.com/caldiaworks/caldiaworks-marketplace
仓库路径:skills/skill-create-workflow
安装命令:
npx skills add https://github.com/caldiaworks/caldiaworks-marketplace --skill skill-create-workflow
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/caldiaworks/caldiaworks-marketplace --skill skill-create-workflow

简介

skill-create-workflow 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。

  • 它能自动分析代码兼容性、依赖关系及协作流程,辅助开发决策。
  • 可通过 npx skills add 命令从 GitHub 安装,具体用法请参考原始 README。
  • 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。
  • 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。

SKILL.md

Skill Development Workflow

Orchestrate the full lifecycle of creating a skill for the CaldiaWorks marketplace.

This workflow combines general-purpose skills (ideation, usdm) with skill-specific tools (skill-creator) to guide the user from a rough idea to a published skill.

Mandatory Execution Rule

Each invoked skill defines its own internal steps (e.g., ideation has Steps 1-6, USDM has Steps 1-5, skill-creator has its own workflow). Every internal step of every invoked skill must be executed in order without exception. Do not skip, merge, or silently omit any step — even if the step seems trivial or the answer seems obvious.

If a step requires user input or confirmation, ask. If a step produces output, present it. If a step offers output format choices, ask the user to choose.

Violating this rule degrades the quality of the output and erodes user trust. There is no acceptable reason to skip a step.

Workflow Steps

Execute these steps in order. Each step produces output that feeds into the next.

If the user already has artifacts from earlier steps (e.g., an idea document or a USDM requirements document), skip the completed steps and start from the appropriate point:

  • User has an idea document → start at Step 2
  • User has a USDM requirements document → start at Step 3
  • User has an implemented skill → start at Step 4

Step 1: Ideation

Invoke the ideation skill to explore and refine the skill idea.

  • Input: User's rough idea or problem description
  • Output: Idea document at .docs/ideas/YYYY-MM-DD-<topic>.md
  • The idea document captures What, Why, Scope, Stakeholders, and Approaches

Ask the user: "Do you want to proceed to requirements definition, or keep this as an idea draft?"

  • If the user wants to proceed → continue to Step 2
  • If the user wants to stop → end the workflow here

Step 2: Requirements Definition

Invoke the usdm skill with the idea document as input.

  • Input: The idea document from Step 1
  • Output: USDM requirements document at .docs/
  • The requirements document decomposes the idea into Requirement → Reason → Description → Specification hierarchy

Ask the user: "Requirements are ready. Proceed to implementation?"

Step 3: Implementation

Invoke the skill-creator skill to implement the skill.

  • Input: The USDM requirements document from Step 2
  • Follow the skill-creator process: understand → plan → initialize → edit → test → iterate
  • Output: A complete skill directory under skills/<skill-name>/

The skill must include:

  • SKILL.md with proper frontmatter
  • Any necessary bundled resources (scripts/, references/, assets/)
  • .claude-plugin/plugin.json

Implementation is not complete until skill-creator's full cycle (including eval and iteration) has been executed. A draft SKILL.md alone does not constitute a completed implementation. The skill-creator defines test cases, runs them, grades expectations, and iterates until stopping criteria are met. All of these steps must be performed.

Step 4: Marketplace Registration

Register the skill in both marketplace manifests.

  1. Verify .claude-plugin/plugin.json exists in the skill directory:
{
  "name": "<skill-name>",
  "version": "0.1.0",
  "description": "<Short description>",
  "skills": ["."]
}
  1. Add the skill path to both root-level manifests:

- .claude-plugin/plugin.json — Add "./skills/<skill-name>" to the skills array - .claude-plugin/marketplace.json — Add "./skills/<skill-name>" to plugins[0].skills array

  1. Bump the version in both files. Both must always have the same version number.
  2. Add the skill to README.md:

- Add the npx skills add install command to the installation section - Add the skill name and description to the Available Skills table

  1. Report completion to the user with the registered skill path.

Step 5: Version Management

If this workflow modified any existing skills (not just the new skill), update their versions.

  1. Check git diff for all files changed during this workflow.
  2. For each skill directory under skills/ that has modifications, bump the version field in its .claude-plugin/plugin.json.
  3. Do not modify files under .claude/ — that directory is managed by the plugin installation system, not by this workflow.

Step 6: Final Verification

Before reporting completion, verify all artifacts are consistent.

  1. Every skill listed in .claude-plugin/marketplace.json plugins[0].skills is also listed in .claude-plugin/plugin.json skills, and vice versa.
  2. Both root manifests have the same version number.
  3. Every skill listed in the manifests exists under skills/<name>/ with a SKILL.md and .claude-plugin/plugin.json.
  4. Every skill listed in the manifests has an entry in README.md (install command + table row).
  5. No unintended files were added to .gitignore — check that existing patterns are not duplicated.

Step 7: Branch Cleanup

After all PRs are merged and the workflow is complete, clean up the working environment.

  1. Switch back to the develop branch and pull the latest changes.
  2. Delete remote feature branches created during this workflow using git push origin --delete <branch>.
  3. Run git fetch --prune to update remote tracking state.
  4. Delete local feature branches created during this workflow whose upstream is gone.
  5. Delete the eval workspace directory if it exists (skills/<skill-name>-workspace/).
  6. Verify the working tree is clean with git status and git branch -a.

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

35.34%
按下载量换算22

Claude

31.76%
按下载量换算20

Cursor

17.01%
按下载量换算11

Gemini CLI

10.02%
按下载量换算6

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

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。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