Token导航 LogoToken导航

Agent Skill · AI 工具

workflow-enforcement(工作流程执行)

workflow-enforcement 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。

workflow-enforcement 是AI 工具类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为只读;许可证为 MIT,第三方扫描有未通过项,按安装量在AI 工具类的 2,536 个 Skill 里排第 631。

总安装 2,671周安装 107GitHub Stars 55下载 865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
MIT

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:workflow-enforcement(工作流程执行)
来源仓库:https://github.com/rysweet/amplihack
仓库路径:skills/workflow-enforcement
安装命令:
npx skills add https://github.com/rysweet/amplihack --skill workflow-enforcement
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。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/rysweet/amplihack --skill workflow-enforcement

这个 Skill 是做什么的

AI 工具类 Skill 用来扩展助手本身的能力,例如接模型、接检索、接外部服务。

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 中整理协作事项。支持围绕仓库状态、代码变更或协作事项进行信息组织与汇总。通过 npx skills add 命令从指定 GitHub 仓库安装,需确认权限与维护状态。

安装前建议核实是否会触发联网、命令执行或文件读写操作。可结合原始 README 进一步了解具体用法和功能边界。

来源仓库内路径:skills/workflow-enforcement

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/rysweet/amplihack --skill workflow-enforcement;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

    它标注为只读,先用只读任务试一次,确认行为符合预期再扩大使用范围。

常见问题

6 条

workflow-enforcement 怎么安装?

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

workflow-enforcement 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。

workflow-enforcement 需要什么权限?

权限级别是只读。该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

workflow-enforcement 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

workflow-enforcement 用的人多吗?

公开数据是累计安装 2,671 次,近一周安装 107 次,下载 865 次,GitHub 55 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。

workflow-enforcement 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

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。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 标注为只读权限,试用阶段不容易误改本地文件。

先想清楚的地方

  • 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AI 工具类任务、又不想每次都手写提示词的人。
  • 只想先试试效果、不希望它改动本地文件的人。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

AI 工具类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
executing-plans

执行计划 · Codex、Claude

1,515,336173,110只读MIT通过
verification-before-completion

完成前的验证 · Codex、Claude

1,319,472173,121权限未标注MIT通过
subagent-driven-development

子 Agent 驱动开发 · Codex、Claude

1,312,416173,138只读MIT通过
copy-editing

复制编辑 · Codex、Claude

1,209,57625,632只读MIT通过
paywall-upgrade-cro

付费专区升级 cro · Codex、Claude

1,030,17625,617只读MIT通过
workflow-enforcement

工作流程执行 · 本页

2,67155只读MIT未通过

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

Workflow Enforcement Skill

Purpose

Guides Claude to complete all workflow steps by:

  1. Reminding about step completion tracking (use TodoWrite or ~/.amplihack/.claude/runtime/workflow_state.yaml)
  2. Emphasizing mandatory steps (0, 10, 16-17) that must not be skipped
  3. Providing visual progress indicator format
  4. Defining expected blocking behavior at checkpoints

Implementation Status: This skill is currently a SPECIFICATION that guides Claude behavior. Actual blocking enforcement requires either:

  • Claude self-compliance (current state)
  • Future: Pre-commit hooks or CI checks (not yet implemented)

The Problem (Issue #1607)

Agents routinely skip mandatory workflow steps, especially:

  • Step 0: Workflow Preparation - Create todos for ALL steps (MANDATORY)
  • Step 10: Pre-commit code review (MANDATORY)
  • Step 16: PR review (MANDATORY)
  • Step 17: Review feedback implementation (MANDATORY)

Root causes:

  • Completion Bias: Agent considers "PR created" as completion instead of "PR merged after review"
  • Context Decay: After heavy implementation, agent loses sight of remaining steps
  • Autonomy Misapplication: Confuses "autonomous implementation" with "skip mandatory process"

Philosophy Alignment and Honest Limitations

The Irony

This skill was created because an agent skipped workflow steps while building a feature. The same cognitive patterns that cause step-skipping can also cause this skill to be ignored. We acknowledge this limitation directly.

What This Skill CAN Do

  • Provide clear guidance patterns for Claude to follow
  • Define standard formats for progress tracking
  • Establish checkpoints where Claude SHOULD pause and validate
  • Integrate with power_steering for session-end verification
  • Document the expected behavior for future agents

What This Skill CANNOT Do

  • Force Claude to read this skill (relies on auto-activation triggers)
  • Prevent Claude from skipping steps if it "decides" to be autonomous
  • Hard-block operations (no pre-commit hooks implemented yet)
  • Catch workflow violations in real-time (only session-end checks exist)

Philosophy Alignment

PrincipleHow This Skill Aligns
Ruthless SimplicitySingle YAML file for state; no complex infrastructure
Zero-BS ImplementationHonest about being guidance, not enforcement
Modular DesignSelf-contained skill with clear integration points
Fail-OpenOn errors, log and continue; never block users

Self-Aware Note: This skill is itself a brick that could be regenerated from its specification. It does not contain executable code - it is documentation that guides Claude's behavior through loaded context.

How to Use This Skill

Step-by-Step Instructions

At Workflow Start (Step 0)
  1. Initialize state tracking (choose one): Option A - TodoWrite (Recommended): Create TodoWrite entries for ALL 22 steps (0-21): - "Step 0: Workflow Preparation - Create todos for ALL steps" - "Step 1: Prepare the Workspace" -... - "Step 21: Ensure PR is Mergeable - TASK COMPLETION" Option B - YAML state file: cp.claude/templates/workflow_state.yaml.template \.claude/runtime/workflow_state.yaml # Edit workflow_id, task_description, started_at
  2. Display initial progress: WORKFLOW PROGRESS [0/22] [.......................] Step 0: Workflow Preparation Mandatory gates: Step 0 (Prep), Step 10 (Review), Step 16 (PR Review), Step 17 (Feedback)
  3. Verify initialization: Confirm 22 todo items exist before proceeding.
At Each Step Completion
  1. Mark step as completed in TodoWrite or YAML
  2. Update current_step to next step number
  3. Display progress indicator
  4. If approaching mandatory step (10, 16, 17), display reminder
At Mandatory Checkpoints

Before Step 15 (Open PR as Draft):

+------------------------------------------------------------------+
|  CHECKPOINT: Pre-PR Validation                                   |
+------------------------------------------------------------------+
|  [ ] Step 10: Pre-commit code review                            |
|      Status: ??? (check your TodoWrite/YAML state)              |
|                                                                  |
|  If Step 10 is NOT completed:                                    |
|  - STOP: Do not proceed to Step 15                              |
|  - ACTION: Invoke reviewer agent for code review                |
|  - ACTION: Invoke security agent for security review            |
|  - Then return here and verify completion                       |
+------------------------------------------------------------------+

Before Step 21 (Ensure Mergeable):

+------------------------------------------------------------------+
|  CHECKPOINT: Final Validation                                    |
+------------------------------------------------------------------+
|  [ ] Step 10: Pre-commit code review                            |
|  [ ] Step 16: PR review                                         |
|  [ ] Step 17: Review feedback implementation                    |
|                                                                  |
|  ALL mandatory steps must be completed before Step 21.          |
|  Check your TodoWrite/YAML state for status.                    |
+------------------------------------------------------------------+
At Workflow End
  1. Verify all steps completed (or explicitly skipped with documented reason)
  2. Final progress: [######################] 22/22 Steps Complete
  3. Delete ~/.amplihack/.claude/runtime/workflow_state.yaml if used
  4. Log completion to workflow_tracker.py

Visual Progress Formats

Standard Progress Bar

WORKFLOW: DEFAULT_WORKFLOW v1.1.0
PROGRESS: [##########............] 10/22 (45%)

Current: Step 10 - Pre-commit code review (MANDATORY)
Next: Step 11 - Incorporate Review Feedback

Detailed Status Display

+======================================================================+
|                    DEFAULT WORKFLOW - Progress                        |
+======================================================================+
| Task: Add authentication feature                                     |
| Session: session_20251125_143022                                     |
| Started: 2025-11-25T14:30:22                                         |
+----------------------------------------------------------------------+
|  Progress: 10/22 steps (45%)                                         |
|  [##########............] 10/22                                       |
+----------------------------------------------------------------------+
|  COMPLETED (10):                                                     |
|    0, 1, 2, 3, 4, 5, 6, 7, 8, 9                                      |
|                                                                      |
|  CURRENT:                                                            |
|    >> Step 10: Pre-commit code review (MANDATORY) <<                 |
|                                                                      |
|  REMAINING (12):                                                     |
|    11, 12, 13, 14, 15, *16*, *17*, 18, 19, 20, 21                   |
|    (* = mandatory)                                                   |
+----------------------------------------------------------------------+
|  MANDATORY GATES:                                                    |
|    [X] Step 0  - Workflow Preparation      COMPLETED                 |
|    [ ] Step 10 - Pre-commit Review         IN PROGRESS               |
|    [ ] Step 16 - PR Review                 PENDING                   |
|    [ ] Step 17 - Feedback Implementation   PENDING                   |
+======================================================================+

Compact Status (For Updates)

[10/22] Step 10 (MANDATORY) | Gates: 0[X] 10[>] 16[ ] 17[ ]

State File Format

Template Location: ~/.amplihack/.claude/templates/workflow_state.yaml.template Active State Location: ~/.amplihack/.claude/runtime/workflow_state.yaml

workflow_id: "session_20251125_143022"
workflow_name: DEFAULT
task_description: "Add authentication feature"
started_at: "2025-11-25T14:30:22"
current_step: 10

steps:
  0: { status: completed, timestamp: "2025-11-25T14:30:22", mandatory: true }
  1: { status: completed, timestamp: "2025-11-25T14:31:05" }
  # ... steps 2-9 ...
  10: { status: in_progress, mandatory: true }
  # ... steps 11-15 ...
  16: { status: pending, mandatory: true }
  17: { status: pending, mandatory: true }
  # ... steps 18-21 ...

mandatory_steps: [0, 10, 16, 17]

checkpoints:
  before_step_15:
    required_steps: [10]
    error_message: "Cannot open PR without completing Step 10"
  before_step_21:
    required_steps: [10, 16, 17]
    error_message: "Cannot mark mergeable without all mandatory reviews"

Integration with Power Steering

Power steering provides session-end verification through the dev_workflow_complete consideration in considerations.yaml.

How It Works

  1. During Session: This skill provides guidance; Claude should track steps in TodoWrite
  2. At Session End: Power steering's _check_dev_workflow_complete analyzes the transcript
  3. Heuristics Used:

- Were architect, builder, reviewer agents invoked? - Were tests executed (Bash tool usage)? - Were git operations performed?

  1. Result: Blocks session completion if workflow evidence is missing

Integration Points

considerations.yaml
==================
- id: dev_workflow_complete
  category: Workflow Process Adherence
  question: Was full DEFAULT_WORKFLOW followed?
  severity: blocker
  checker: _check_dev_workflow_complete
  applicable_session_types: ["DEVELOPMENT"]

Current Limitation: Power steering checks for evidence of workflow (agent usage, test runs) but does NOT read workflow_state.yaml. This is a future enhancement opportunity.

Proposed Enhancement (Not Yet Implemented)

# In power_steering_checker.py (future)
def _check_dev_workflow_complete(self, transcript, session_id):
    # Check for workflow_state.yaml
    state_file = Path(".claude/runtime/workflow_state.yaml")
    if state_file.exists():
        state = yaml.safe_load(state_file.read_text())
        mandatory = state.get("mandatory_steps", [0, 10, 16, 17])
        steps = state.get("steps", {})

        for step_num in mandatory:
            step = steps.get(step_num, {})
            if step.get("status") != "completed":
                return False  # Mandatory step incomplete

        return True  # All mandatory steps completed

    # Fall back to heuristic analysis
    return self._check_workflow_heuristics(transcript)

Integration with TodoWrite

When using TodoWrite, ensure step numbers match workflow tracking:

TodoWrite(todos=[
    {"content": "Step 0: Workflow Preparation - Create todos for ALL steps", "status": "completed", "activeForm": "Creating step todos"},
    {"content": "Step 10: Pre-commit code review - MANDATORY", "status": "in_progress", "activeForm": "Reviewing code"},
    {"content": "Step 16: PR review - MANDATORY", "status": "pending", "activeForm": "Reviewing PR"},
    {"content": "Step 17: Implement review feedback - MANDATORY", "status": "pending", "activeForm": "Implementing feedback"},
])

Key Point: TodoWrite is the primary tracking mechanism. YAML state file is optional for additional persistence and structured validation.

Blocking Behavior (Self-Compliance Pattern)

When mandatory steps are skipped, Claude SHOULD display and follow this pattern:

+======================================================================+
|                    WORKFLOW ENFORCEMENT: BLOCKED                      |
+======================================================================+
|                                                                      |
|  Cannot proceed to Step 15 (Open PR as Draft).                       |
|                                                                      |
|  MISSING MANDATORY STEP:                                             |
|    Step 10: Pre-commit code review                                   |
|                                                                      |
|  WHY THIS MATTERS:                                                   |
|    - Code review catches bugs before they reach CI                   |
|    - Philosophy compliance ensures quality                           |
|    - Issue #1607 identified this as a recurring problem              |
|                                                                      |
|  ACTION REQUIRED:                                                    |
|    1. Invoke reviewer agent for comprehensive code review            |
|    2. Invoke security agent for security review                      |
|    3. Mark Step 10 as completed in TodoWrite                         |
|    4. Then proceed to Step 15                                        |
|                                                                      |
+======================================================================+

Error Recovery

If state file is missing or corrupt:

  1. Check TodoWrite for current step information
  2. Reconstruct state from conversation context
  3. If unrecoverable, prompt user to confirm current step
  4. Always fail-open: continue with warning rather than blocking

Future Work: Path to Hard Enforcement

Phase 1: Current State (Guidance Only)

  • Skill provides documentation and patterns
  • Claude self-compliance determines effectiveness
  • Power steering provides session-end warnings

Phase 2: Enhanced Power Steering (Planned)

  • Power steering reads workflow_state.yaml
  • Real-time validation at checkpoint steps
  • More granular dev_workflow_complete checks

Phase 3: Pre-Commit Integration (Future)

# .pre-commit-config.yaml (future)
- repo: local
  hooks:
    - id: workflow-validation
      name: Validate workflow completion
      entry: python .claude/tools/amplihack/validate_workflow.py
      language: system
      pass_filenames: false

Phase 4: CI Gate (Future)

# .github/workflows/ci.yml (future)
- name: Validate Workflow Compliance
  run: |
    python -c "
    from pathlib import Path
    import yaml
    state = yaml.safe_load(Path('.claude/runtime/workflow_state.yaml').read_text())
    mandatory = state.get('mandatory_steps', [0, 10, 16, 17])
    for step in mandatory:
        if state['steps'].get(step, {}).get('status') != 'completed':
            print(f'FAIL: Mandatory step {step} not completed')
            exit(1)
    print('PASS: All mandatory workflow steps completed')
    "

Related Components

  • DEFAULT_WORKFLOW.md: Canonical workflow definition with Step 0 guidance
  • templates/workflow_state.yaml.template: Template for structured state tracking
  • workflow_tracker.py: Historical logging (JSONL format)
  • power_steering_checker.py: Session-end enforcement (transcript-based)
  • considerations.yaml: dev_workflow_complete consideration definition

平台分布

Claude Code

27.64%
按下载量换算239

OpenCode

22.6%
按下载量换算195

Antigravity

18.04%
按下载量换算156

Gemini CLI

12.27%
按下载量换算106

windsurf

7.19%
按下载量换算62

Cursor

3.42%
按下载量换算30

安全审计

Gen Agent Trust Hub

未通过

Socket

通过

Snyk

通过

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

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。

继续浏览同类 Skills