Planning And Task Breakdown

Planning and Task Breakdown

Overview

Break a specification into atomic, independently verifiable tasks with clear acceptance criteria, dependency ordering, and time estimates. This is the canonical Plan-phase skill.

When to Use

• After a spec is approved and before implementation begins
• When a feature feels too large to implement in one pass
• When multiple people or agents need to work on related changes
• Before using incremental-implementation or test-driven-development

Task Hierarchy

Epic (the full feature)
└── Phase (logical grouping, ~1 day of work)
    └── Task (atomic unit, ~30 min)
        └── Subtask (only if task is still too large)

Task Requirements

Every atomic task must include:

Process

Step 1: Read the Spec

• Extract all user stories and acceptance criteria
• Identify non-functional requirements
• Note constraints and boundaries

Step 2: Map Stories to Tasks

For each user story, break it into implementation tasks:

• One task per cohesive change
• Each task should be completable in under an hour
• Tasks should be verifiable independently when possible

Step 3: Order by Dependency

• Identify blocking relationships
• Group into phases where tasks within a phase can run in parallel
• Ensure each phase produces a buildable, testable state

Step 4: Estimate and Validate

• Assign time estimates
• Validate that the total plan covers all spec acceptance criteria
• Check that no task is too large or too vague

Step 5: Create Tasks

Primary: Solo (if available)

Check if Solo is available:

# Check if Solo directory exists and CLI is available
test -d ~/Github/Solo && which solo >/dev/null 2>&1

If Solo is available, create each task using the solo task create command:

solo task create \
  --title "Task title" \
  --description "Full task description with acceptance criteria and files" \
  --priority medium \
  --deps T-1,T-2 \
  --tags feature-name,phase-1 \
  --json

Solo task fields:

Solo returns the assigned task ID (e.g., T-142). Use this ID for dependency references in subsequent tasks.

Fallback: File-based tasks (if Solo is not available)

If Solo is not available, create individual task files:

specs/{feature}/tasks/
├── T-001.md
├── T-002.md
└── ...

Each task file follows this format:

# T-001: [Task Title]

**Depends On:** None
**Priority:** medium
**Estimated Time:** 30 min
**Complexity:** Low

## Acceptance Criteria
- [ ] Criterion 1
- [ ] Criterion 2

## Files
- `path/to/file.ts` - Create/modify description

## Spec Reference
- US-001: AC-2

Common Rationalizations

Verification

• Every spec acceptance criterion is covered by at least one task
• Every task has all required fields
• Dependency graph has no cycles
• Each phase produces a buildable, testable state
• Tasks created in Solo (if available) OR saved as specs/{feature}/tasks/*.md (fallback)
• Solo task IDs recorded for dependency tracking (if using Solo)

Anti-Rationalization Table