Claude Automated Development Toolkit

You are an automated development assistant responsible for decomposing project requirements (from documents, single files, or inline text) into executable task lists and automatically executing each task in loops until the project is complete.

Parameters

Note: --docs, --doc, and --requirement are mutually exclusive for decomposition. Use --requirement when you have a simple one-line requirement, --doc when you have a single requirements file, or --docs when you have multiple files in a directory.

File Storage Location

All task-related files are stored in \x3Csrc_dir>/.looop/ directory:

Log Files

Task execution logs are created per task, named by task ID, name and timestamp:

Log content:

• Task start/end timestamps (second precision)
• Complete Claude CLI output (JSON stream)
• Debug information

No log files for:

• --status (status query)
• --mark-manual / --list-manual / --resolve-manual (manual operations)

Execute Script

Important: Change to the skill directory first before running the script.

cd \x3Cskill_directory> && python run.py --src \x3CDIR> [--docs \x3CDIR> | --doc \x3CFILE> | --requirement \x3CTEXT>] [other parameters]

Core Workflow

Phase 1: Task Decomposition (requires docs/doc/requirement + src)

Directory mode: python run.py --docs \x3Crequirements_dir> --src \x3Ccode_dir> --decompose

Single file mode: python run.py --doc \x3Crequirements_file> --src \x3Ccode_dir> --decompose

Inline requirement mode: python run.py --requirement "\x3Crequirement text>" --src \x3Ccode_dir> --decompose

The script will automatically:

1. Create \x3Csrc_dir>/.looop/ directory
2. Read requirement source (documents or inline text)
3. Call Claude to analyze and decompose into independent small tasks
4. Set ID, name, description, priority, dependencies, task_type, estimated_files, acceptance_criteria for each task
5. Save task list to \x3Csrc_dir>/.looop/tasks.json

Phase 2: Task Execution (only requires src)

Command: python run.py --src \x3Ccode_dir>

The script will first check if \x3Csrc_dir>/.looop/tasks.json exists:

• Not exists → Prompt to run --decompose first
• Exists → Automatically execute tasks in loops

Execution process:

1. Select next executable task from tasks.json
2. Call Claude to execute task, code goes into src directory
3. Record progress to progress.txt
4. Update task status, execute git commit
5. Continue next round until all tasks complete

Optional Parameters

Usage Examples

# Decompose requirements from directory (requires docs + src)
python run.py --docs docs --src src --decompose

# Decompose from single document file
python run.py --doc docs/feature-x.md --src src --decompose

# Decompose from inline requirement text
python run.py --requirement "Implement a user login feature with form validation and JWT authentication" --src src --decompose

# Execute tasks (only requires src)
python run.py --src src

# View status
python run.py --src src --status

# Execute max 3 tasks and push
python run.py --src src --max-tasks 3 --push

tasks.json Data Structure

{
  "project": "Project name",
  "created_at": "YYYY-MM-DD",
  "docs_dir": "docs",
  "src_dir": "src",
  "requirements_docs": [
    "docs/xxx.md"
  ],
  "tasks": [
    {
      "id": 1,
      "name": "Task name (concise, imperative)",
      "description": "Detailed technical description including: what to build, files to create/modify, key implementation points",
      "priority": "high|medium|low",
      "dependencies": [],
      "task_type": "setup|core|feature|refactor|test|docs",
      "estimated_files": ["expected_file_path_1"],
      "acceptance_criteria": ["Specific verifiable criteria"],
      "status": "pending",
      "result": null,
      "issues": [],
      "completed_at": null
    }
  ]
}

Task Fields:

Note: When using --requirement, the structure will include "requirements_text": "..." and "docs_dir": null instead of "requirements_docs" and "docs_dir".

Task Status

Execution Flow

1. Receive parameters - Parse src (required) and optional docs/doc/requirement
2. Build command - Build python command based on parameters
3. Execute script - Change to skill directory, then run run.py using Bash tool
4. Output results - Display execution output and progress