How to manage overall project workflow and orchestrate skill execution. Make sure to use this skill whenever the user mentions workflow management, project progress, advancing to the next stage, or needs to coordinate multiple skills. Also trigger when initializing new projects, starting new features, fixing bugs, or updating documentation. Works closely with brainstorming, project-structure, project-docs, writing-plans, project-task, executing-plans, and other skills.
Project workflow orchestrates skill execution across stages, providing structured tracking and progress reporting.
| Scenario | Flow |
|----------|------|
| project_init | 1. brainstorming → 2. project-structure → 3. project-docs → 4. writing-plans → 5. [execute] → 6. project-docs |
| feature_dev | 1. brainstorming → 2. project-docs → 3. writing-plans → 4. [execute] → 5. requesting-code-review → 6. finishing-a-development-branch |
| bug_fix | 1. brainstorming → 2. project-docs → 3. systematic-debugging → 4. [execute] → 5. verification-before-completion |
| docs_update | 1. brainstorming → 2. project-docs → 3. [execute] |
| refactor | 1. analyze existing code → 2. project_init → 3. feature_dev |
| tech_research | 1. brainstorming → 2. [execute] → 3. project-docs |
Workflow principles:
brainstorming — User participation pointproject-docs — Code-doc alignment check[execute] = Choose between executing-plans or subagent-driven-development (see below)
Use dispatching-parallel-agents for multiple independent tasks:
| Use Case | Condition | |----------|-----------| | Multiple modules | Different subsystems | | Multiple bugs | Different root causes, no shared state | | Multiple features | No dependencies |
Requirements: No shared state, no sequential dependencies, each task independent.
| Mode | Skill | Session | Review | Best For |
|------|-------|---------|--------|----------|
| Parallel | executing-plans | New worktree | After all tasks | Major features, long work |
| Same | subagent-driven-development | Current | After each task | Independent tasks, quick fixes |
Decision:
subagent-driven-developmentexecuting-plansTraditional AI programming mindset:
I ask → AI answers → I ask again → AI answers again
Opencode workflow mindset:
I give instruction → AI breaks down tasks → Multiple Agents collaborate → Automatically complete
When you master "orchestration" rather than "conversation", you become an AI team manager rather than an AI user.
| Complexity | Indicators | Recommended Approach | |------------|------------|---------------------| | Low | "typo", "rename", "simple", "minor" | Quick-fix (skip workflow) | | Medium | Standard feature/bug fixes | User choice | | High | "architecture", "refactor", "migrate", "async" | Full workflow |
| Scenario | ID | Use Case |
|----------|-----|----------|
| Project Initialization | project_init | New project, repository setup |
| Feature Development | feature_dev | Adding new features, modules |
| Bug Fix | bug_fix | Fixing bugs, issue resolution |
| Documentation Update | docs_update | Design changes, doc sync |
| Code Refactoring | refactor | Restructuring existing code |
| Technical Research | tech_research | Technology selection, POC |
project_init)| Stage | Skill | Output |
|-------|-------|--------|
| 1 | brainstorming | Project design |
| 2 | project-structure | src/, tests/, docs/, scripts/ |
| 3 | project-docs | docs/design/README.md |
| 4 | writing-plans | docs/plans/YYYY-MM-DD-init-plan.md |
| 5 | [execute] | Code implementation |
| 6 | project-docs | Code-doc alignment check |
[execute] = executing-plans or subagent-driven-development (see Execution Mode Selection above)
feature_dev)| Stage | Skill | Output |
|-------|-------|--------|
| 1 | brainstorming | Feature design |
| 2 | project-docs | docs/design/<feature>/README.md |
| 3 | writing-plans | docs/plans/YYYY-MM-DD-feature-plan.md |
| 4 | [execute] | Code implementation |
| 5 | requesting-code-review | Code review report |
| 6 | finishing-a-development-branch | Branch merged |
bug_fix)| Stage | Skill | Output | |-------|-------|--------| | 1 | brainstorming | Problem analysis (design issue vs code issue) | | 2 | systematic-debugging | Root cause analysis | | 3 | [execute] | Bug fix code | | 4 | verification-before-completion | Verification report |
Issue classification:
docs_update)| Stage | Skill | Output | |-------|-------|--------| | 1 | brainstorming | Update scope definition | | 2 | project-docs | Updated design docs | | 3 | [execute] | Documentation changes |
refactor)| Stage | Skill | Output | |-------|-------|--------| | 1 | brainstorming | Refactoring analysis | | 2 | [execute] | Refactored code | | 3 | verification-before-completion | Tests pass verification | | 4 | requesting-code-review | Code quality review | | 5 | project-docs | Updated design docs |
tech_research)| Stage | Skill | Output |
|-------|-------|--------|
| 1 | brainstorming | Option analysis, recommendation |
| 2 | [execute] | Prototype implementation |
| 3 | project-docs | Research report (docs/research/<topic>.md) |
Workflow stages automatically trigger corresponding skills:
| Stage | Skill | Purpose |
|-------|-------|---------|
| Design phase | brainstorming | Design refinement with user participation |
| Structure phase | project-structure | Project scaffolding |
| Docs phase | project-docs | Design documentation sync |
| Planning phase | writing-plans | Implementation planning |
| Execution phase | executing-plans / subagent-driven-development | Code execution |
| Debug phase | systematic-debugging | Root cause analysis |
| Review phase | requesting-code-review | Code quality check |
| Completion phase | finishing-a-development-branch | Branch merge and cleanup |
| Verification phase | verification-before-completion | Fix verification |
Skill invocation:
skill tool to load skill contentAlways start with design. Use brainstorming skill to:
Break down tasks. Use writing-plans skill to:
Never skip review. Use requesting-code-review skill before completion:
Define constraints clearly:
Create AGENTS.md in project root:
# 项目规范
## 技术栈
- 前端：React 18 + TypeScript + Vite
- 后端：Node.js + Express + PostgreSQL
- 部署：Docker + AWS
## 编码规范
- 使用 ESLint + Prettier
- 组件使用函数式 + Hooks
- 状态管理使用 Zustand
- API 调用使用 React Query
## 目录结构
- `/src/components` - 可复用组件
- `/src/features` - 功能模块
- `/src/hooks` - 自定义 Hooks
- `/src/utils` - 工具函数
## 命名规范
- 组件：PascalCase
- 函数：camelCase
- 常量：UPPER_SNAKE_CASE
- 文件：kebab-case
| Pitfall | Wrong | Right | |---------|-------|-------| | One-shot large tasks | "帮我实现一个完整的电商系统" | Break into phases: design → modules → integration | | No clear boundaries | "优化这个模块的性能" | Define constraints: no API changes, compatibility, tests | | Skip code review | 实现 → 直接提交 | 实现 → 审查 → 修复 → 提交 | | Ignore testing | 功能实现完就算完成 | 实现 → 单元测试 → 集成测试 → 手动测试 |
Skip workflow when:
Cancel workflow and use quick-fix for the actual change.
| Task Type | Scenario |
|-----------|----------|
| Fixing bugs | bug_fix |
| Adding features | feature_dev |
| Updating docs | docs_update |
| New project | project_init |
| Refactoring code | refactor |
| Technology research | tech_research |
Good workflow principles:
Google Workspace CLI for Gmail, Calendar, Drive, Contacts, Sheets, and Docs.
Manage Apple Notes via the `memo` CLI on macOS (create, view, edit, delete, search, move, and export notes). Use when a user asks OpenClaw to add a note, list notes, search notes, or manage note folders.
Work with Obsidian vaults (plain Markdown notes) and automate via obsidian-cli.
Use when you need to control Slack from OpenClaw via the slack tool, including reacting to messages or pinning/unpinning items in Slack channels or DMs.
Manage Apple Reminders via remindctl CLI (list, add, edit, complete, delete). Supports lists, date filters, and JSON/plain output.
Manage Trello boards, lists, and cards via the Trello REST API.
Category:productivity