kurko/working-off-of-todo-files

Overview

This skill enables Claude Code to work systematically with ai-notes/**/*/todo.{md,txt} files in your codebase. The agent focuses on completing one task at a time with precision, breaking complex tasks into subtasks mentally before execution.

Core Principles

1. Single Task Focus: Work on ONE task at a time until completion or user intervention
2. Mental Decomposition: Break tasks into subtasks in memory before starting
3. Progressive Status: Use clear markers for task progress
4. User Control: Only advance to next task when explicitly instructed
5. Context Awareness: Understand the broader project while maintaining task focus

Task Format Specification

Basic Structure

## Context
Brief overview of the project goals and current state

## Tasks

- [ ] Task 1: Clear task description
    Additional context or requirements indented by 4 spaces
    Can include multiple lines of detail
    
- [ ] Task 2: Another task
    Context and requirements
    
- [/] Task 3: Currently in progress
    This task is being worked on
    Progress notes can be added here
    
- [x] Task 4: Completed task
    This task has been finished
    Completion notes or results

Task States

• - [ ] Pending: Task not yet started
• - [/] In Progress: Currently working on this task
• - [x] Completed: Task finished successfully

Working Protocol

1. Initial Assessment

When starting work on a todo.md file:

1. Read the entire file to understand project context
2. Identify the next uncompleted task (first `- [ ]`, unless instructed otherwise)
3. Mentally decompose the task into subtasks
4. Announce which task you'll be working on

2. Task Execution Process

Before Starting

• Identify the task clearly: "I'll work on: [Task description]"
• Generate mental subtasks (do not write these to file):
  • Analyze requirements
  • Identify dependencies
  • Plan implementation steps
  • Define success criteria
• Mark task as in-progress: change - [ ] to - [/]

During Execution

• Work through mental subtasks systematically
• Add progress notes under the task (indented)
• If encountering blockers, document them
• Keep changes focused on the current task only

After Completion

• Verify all requirements are met
• Run relevant tests/specs
• Mark task complete: change - [/] to - [x]
• Add completion notes if valuable
• Stop and wait for user confirmation to continue

3. Communication Protocol

Always communicate:

• Which task you're starting
• Key decisions or assumptions
• Any blockers encountered
• When task is complete
• Request for permission to change plan if needed. Explain why and trade-offs.
• Request for permission to continue

Example interaction:

Agent: "I've identified Task 2: 'Implement user authentication'. I'll start by setting up the authentication middleware. Marking as in progress."

[Works on task]

Agent: "Task 2 is complete. The authentication system is now functional with JWT tokens. Should I proceed to Task 3 for adding a password reset functionality?"

User: "Yes, continue"

Agent: "Starting Task 3: 'Add password reset functionality'..."

Mental Subtask Generation

For each task, generate (but don't write) subtasks like:

Example: "Implement API endpoint"

Mental subtasks:

1. Define route structure
2. Write e2e tests (if applicable) to settle the expectations
3. Create controller function
4. Add input validation
5. Implement business logic
6. Add error handling
7. Write tests
8. Consider simplifications, optimizations, and readability.
9. Update documentation

Example: "Fix bug in payment processing"

Mental subtasks:

1. Reproduce the bug
2. Write a test that fails due to the bug
3. Identify root cause
4. Review related code
5. Implement fix
6. Test edge cases
7. Run test that reproduced the bug
8. Verify no regression with existing tests
9. Document the fix

File Discovery and Management

Locating Todo Files

# Find all todo.md files in ai-notes directories
find . -path "*/ai-notes/*" -name "todo.md" -type f

# List with context
find . -path "*/ai-notes/*" -name "todo.md" -exec echo "Found: {}" \; -exec head -n 5 {} \;

Safe File Operations

• Always backup before major changes
• Use atomic operations when possible
• Preserve file formatting and structure
• Maintain consistent indentation (4 spaces for task details)

Task Dependencies and Context

Understanding Dependencies

Before starting a task, check for:

• References to other tasks
• Required prerequisites
• Related code files mentioned
• External dependencies

Maintaining Context

• Keep track of completed tasks for reference
• Note patterns or decisions from previous tasks
• Build on work from completed tasks
• Document assumptions for future tasks

Error Handling

Common Scenarios

1. Ambiguous task description

   • Ask for clarification
   • Document interpretation

2. Blocked by external dependency

   • Document the blocker
   • Suggest alternatives or workarounds
   • Move to next task only if instructed

3. Task too large

   • Suggest breaking into multiple tasks
   • Complete what's reasonable
   • Document remaining work

Best Practices

Do's

• ✅ Focus completely on one task
• ✅ Think through subtasks before starting
• ✅ Update status markers promptly
• ✅ Add helpful progress notes
• ✅ Verify completion thoroughly
• ✅ Wait for user permission between tasks
• ✅ Maintain clean, readable formatting

Don'ts

• ❌ Jump between multiple tasks
• ❌ Skip marking progress states
• ❌ Add visible subtasks to the file (keep them mental)
• ❌ Continue to next task without permission
• ❌ Modify completed tasks unnecessarily
• ❌ Lose track of the big picture

Advanced Features

Task Annotations

You may encounter special annotations:

- [ ] Task with priority !high
- [ ] Task with deadline @2024-12-31
- [ ] Task with assignee @alex
- [ ] Task with tag #backend

Respect these but don't require them.

• If you see !high, prioritize if user allows

Progress Tracking

When requested, provide summary:

Total Tasks: X
Completed: Y (Z%)
In Progress: 1
Remaining: W
Current Focus: [Task description]

Integration Notes

• Compatible with version control (git-friendly)
• Works alongside issue tracking systems
• Can reference external tickets or PRs
• Supports cross-referencing between projects

Example Session

# Initial State
- [ ] Set up database schema
- [ ] Create API endpoints
- [ ] Add authentication

# After starting Task 1
- [/] Set up database schema
    Creating tables for users, sessions, and permissions
    Using PostgreSQL with migrations
    
# After completing Task 1
- [x] Set up database schema
    Completed: 5 tables created, migrations ready
- [ ] Create API endpoints
- [ ] Add authentication

# User says "continue"
- [x] Set up database schema
    Completed: 5 tables created, migrations ready
- [/] Create API endpoints
    Working on RESTful endpoints for user management
- [ ] Add authentication

Troubleshooting

Issue: Task unclear

Solution: Ask for clarification, document assumptions

Issue: Multiple todo.md files

Solution: Ask which project to focus on or work through them systematically

Issue: Conflicting priorities

Solution: Follow file order unless user specifies otherwise

Issue: Task partially complete from before

Solution: Assess current state, complete remaining work

Remember

• Quality over speed
• One task, done well, is better than multiple tasks done poorly
• The user is in control of pacing
• Document your work for future reference
• Think in subtasks, but execute as one cohesive task