AgentSkillsCN

discipline-refactor

按学科/领域重新组织任意仓库,构建模块化结构。将代码存放于 src/<pkg>/modules/ 目录,文档则置于 <folder>/modules/ 中。统筹 Explore、Plan 以及通用型子代理,可独立运行。

SKILL.md
--- frontmatter
name: discipline-refactor
description: Reorganize any repository by discipline/domain with module-based
  structure. Code in src/<pkg>/modules/, documents in <folder>/modules/.
  Orchestrates Explore, Plan, and general-purpose subagents. Works standalone.
version: 2.0.0
category: refactoring
triggers:
  - "refactor by discipline"
  - "organize by domain"
  - "restructure repository"
  - "module-based organization"
  - "discipline-based organization"
prerequisites: none
standalone: true
calls_skills:
  - skill-creator
  - git-sync-manager
  - parallel-batch-executor
calls_subagents:
  - Explore
  - Plan
  - general-purpose
  - Bash

Discipline-Based Refactor (v2.0)

Reorganize any repository to module-based, discipline-aligned structure.

Core Principles

  1. All folders are module-based - Every directory uses modules/<discipline>/ pattern
  2. Code lives in modules - src/<package>/modules/<discipline>/
  3. Documents mirror code - docs/modules/<discipline>/, specs/modules/<discipline>/
  4. Disciplines are consistent - Same discipline names across all folders

Target Repository Structure

code
<repo>/
├── src/<package_name>/
│   └── modules/
│       ├── _core/              # Shared utilities
│       ├── <discipline-1>/     # Domain module
│       │   ├── __init__.py
│       │   ├── models.py
│       │   ├── services.py
│       │   └── utils.py
│       └── <discipline-2>/
│
├── tests/
│   └── modules/
│       ├── _core/
│       ├── <discipline-1>/
│       └── <discipline-2>/
│
├── docs/
│   └── modules/
│       ├── _core/
│       ├── <discipline-1>/
│       └── <discipline-2>/
│
├── specs/
│   └── modules/
│       ├── _core/
│       ├── <discipline-1>/
│       └── <discipline-2>/
│
├── data/
│   └── modules/
│       ├── <discipline-1>/
│       └── <discipline-2>/
│
├── logs/
│   └── modules/
│       ├── <discipline-1>/
│       └── <discipline-2>/
│
├── .claude/
│   ├── skills/
│   │   ├── _core/
│   │   ├── <discipline-1>/
│   │   └── <discipline-2>/
│   └── CLAUDE.md
│
└── pyproject.toml / package.json

Orchestration Architecture

code
┌─────────────────────────────────────────────────────────────┐
│  ORCHESTRATOR: discipline-refactor skill                    │
│  (Stays lean, delegates all execution)                      │
└─────────────────────────────────────────────────────────────┘
                            │
        ┌───────────────────┼───────────────────┐
        ▼                   ▼                   ▼
┌───────────────┐   ┌───────────────┐   ┌───────────────┐
│ Phase 1       │   │ Phase 2       │   │ Phase 3       │
│ ANALYSIS      │   │ PLANNING      │   │ EXECUTION     │
│ Explore       │   │ Plan          │   │ general-purpose│
│               │   │ + skill-creator│  │ + git-sync-mgr│
└───────────────┘   └───────────────┘   └───────────────┘
                            │
                            ▼
                ┌───────────────────────┐
                │ Phase 4: VALIDATION   │
                │ Bash (run tests)      │
                └───────────────────────┘

Phase 1: Analysis

Spawn: Task with subagent_type=Explore

Prompt:

code
Analyze the repository for discipline-based, module-based refactoring:

1. Identify package name:
   - Check pyproject.toml [project.name] or [tool.poetry.name]
   - Check package.json name
   - Check existing src/<name>/ structure
   - Derive from repo name if not found

2. Scan ALL top-level directories:
   - src/ - code structure
   - tests/ - test organization
   - docs/ - documentation structure
   - specs/ - specifications
   - data/ - data files
   - logs/ - log files
   - .claude/skills/ - skill organization

3. Identify disciplines from existing code:
   - What domain modules exist?
   - What functional areas are present?
   - Map existing directories to discipline names

4. Check for existing modules/ patterns:
   - Already have src/<pkg>/modules/?
   - Already have tests/modules/?
   - What's the current organization level?

5. Output discipline mapping:
   - Suggested disciplines (use consistent names)
   - Current path → new module path for each folder
   - Package name to use

Report in structured format for Phase 2.

Phase 2: Planning

Spawn: Task with subagent_type=Plan

Call: @skill-creator for new discipline skills

Prompt:

code
Create module-based migration plan based on Phase 1 analysis.

Context: {analysis_results}

Create plan for ALL directories:

1. Package identification:
   - Confirm package name: <package_name>
   - Confirm disciplines: [_core, discipline-1, discipline-2, ...]

2. Code structure (src/):
   BEFORE: src/<package>/feature1/, src/<package>/feature2/
   AFTER:  src/<package>/modules/_core/, src/<package>/modules/<discipline>/

3. Test structure:
   BEFORE: tests/test_feature1.py, tests/integration/
   AFTER:  tests/modules/_core/, tests/modules/<discipline>/

4. Documentation structure:
   BEFORE: docs/api/, docs/guides/
   AFTER:  docs/modules/_core/, docs/modules/<discipline>/

5. Specs structure:
   BEFORE: specs/feature-spec.md, specs/plan.md
   AFTER:  specs/modules/<discipline>/, specs/templates/

6. Data structure (if exists):
   BEFORE: data/raw/, data/processed/
   AFTER:  data/modules/<discipline>/

7. Skills structure:
   BEFORE: .claude/skills/category/
   AFTER:  .claude/skills/<discipline>/

8. Import updates:
   - from <package>.feature → from <package>.modules.<discipline>
   - Update __init__.py files
   - Update relative imports

9. Create README.md for each module with:
   - Module purpose
   - Key files
   - Dependencies on other modules

Output executable task list for Phase 3.

Phase 3: Execution

Spawn: Task with subagent_type=general-purpose

Call: @git-sync-manager, @parallel-batch-executor

Prompt:

code
Execute module-based refactoring:

Migration Plan: {migration_plan}
Package Name: {package_name}
Disciplines: {disciplines}

Execute in order:

1. BACKUP:
   git tag pre-module-refactor-$(date +%Y%m%d)

2. CREATE MODULE STRUCTURE:
   # For each discipline in [_core, discipline-1, ...]:
   mkdir -p src/<package>/modules/<discipline>
   mkdir -p tests/modules/<discipline>
   mkdir -p docs/modules/<discipline>
   mkdir -p specs/modules/<discipline>
   mkdir -p data/modules/<discipline>  # if data/ exists
   mkdir -p .claude/skills/<discipline>

3. MOVE CODE:
   # Move source files to appropriate modules
   # Update __init__.py in each module
   # Preserve git history with git mv

4. MOVE TESTS:
   # Move test files to mirror source structure
   # Update test imports

5. MOVE DOCS:
   # Move documentation to module folders
   # Update internal links

6. MOVE SPECS:
   # Move specifications to module folders
   # Keep templates/ at top level

7. UPDATE IMPORTS:
   # Search and replace import paths
   # from <pkg>.old_path → from <pkg>.modules.<discipline>
   # Update conftest.py
   # Update pyproject.toml entry points

8. UPDATE CONFIGS:
   # pyproject.toml: packages, entry points
   # Update CI/CD paths
   # Update Makefile/scripts

9. CREATE MODULE README:
   # Add README.md to each module explaining purpose

Report progress. Stop on failure.

Phase 4: Validation

Spawn: Task with subagent_type=Bash

Prompt:

code
Validate module-based refactoring:

1. Structure validation:
   - [ ] src/<package>/modules/ exists with disciplines
   - [ ] tests/modules/ mirrors src structure
   - [ ] docs/modules/ exists
   - [ ] specs/modules/ exists
   - [ ] Each module has __init__.py (Python) or index (JS)
   - [ ] Each module has README.md

2. Import validation:
   - [ ] No old import paths remain
   - [ ] All imports resolve
   - [ ] Circular imports checked

3. Test validation:
   - [ ] pytest / npm test passes
   - [ ] Coverage maintained

4. Documentation validation:
   - [ ] Internal links work
   - [ ] Module READMEs complete

Generate report. On failure, suggest:
git checkout pre-module-refactor-{date}

Module-Based Folder Mapping

Standard Folders → Module Structure

FolderModule PatternExample
src/<pkg>/src/<pkg>/modules/<discipline>/src/myapp/modules/data/
tests/tests/modules/<discipline>/tests/modules/data/
docs/docs/modules/<discipline>/docs/modules/data/
specs/specs/modules/<discipline>/specs/modules/data/
data/data/modules/<discipline>/data/modules/ingestion/
logs/logs/modules/<discipline>/logs/modules/api/
.claude/skills/.claude/skills/<discipline>/.claude/skills/data/

Exceptions (Keep Flat)

FolderReason
specs/templates/Shared templates
docs/assets/Shared images/files
.claude/state/Runtime state
scripts/Build/deploy scripts
config/Configuration files

Discipline Taxonomy

Standard disciplines (customize per repo):

DisciplinePurposeContains
_coreShared utilitiesBase classes, logging, config, utils
engineeringDomain expertiseSimulation, analysis, calculations
dataData handlingETL, storage, visualization
apiExternal interfacesREST, GraphQL, webhooks
automationWorkflowsCI/CD, scripts, jobs
_internalRepo-specificMeta tools, guidelines

How to Identify Disciplines

  1. What are the main "nouns" in this repo?
  2. What expertise areas does the code serve?
  3. What would you call the team that owns each part?
  4. Use consistent names across src/, tests/, docs/, specs/

Example Transformations

Python Package

Before:

code
mypackage/
├── src/mypackage/
│   ├── utils.py
│   ├── models.py
│   ├── api/
│   └── data/
├── tests/
│   ├── test_utils.py
│   └── test_api.py
└── docs/
    ├── api.md
    └── data.md

After:

code
mypackage/
├── src/mypackage/
│   └── modules/
│       ├── _core/
│       │   ├── __init__.py
│       │   └── utils.py
│       ├── api/
│       │   ├── __init__.py
│       │   └── routes.py
│       └── data/
│           ├── __init__.py
│           └── models.py
├── tests/
│   └── modules/
│       ├── _core/
│       │   └── test_utils.py
│       └── api/
│           └── test_routes.py
├── docs/
│   └── modules/
│       ├── api/
│       │   └── README.md
│       └── data/
│           └── README.md
└── specs/
    └── modules/
        └── api/
            └── api-spec.md

Import Changes

python
# Before
from mypackage.utils import helper
from mypackage.api.routes import router

# After
from mypackage.modules._core.utils import helper
from mypackage.modules.api.routes import router

Rollback

bash
git checkout pre-module-refactor-{date}

Verification Checklist

  • All code in src/<pkg>/modules/<discipline>/
  • All tests in tests/modules/<discipline>/
  • All docs in docs/modules/<discipline>/
  • All specs in specs/modules/<discipline>/
  • Consistent discipline names across all folders
  • Each module has __init__.py and README.md
  • All imports updated
  • Tests pass
  • No orphaned files outside modules/