Agent Skills: Tasks Generator

Tasks Generator

Generate structured, actionable task roadmaps from project specifications (PRD, Tech Specs, UI/UX Specs).

Workflow

1. Gather inputs - Locate and read PRD, Tech Specs, and UI/UX specs
2. Analyze specifications - Extract features, requirements, and dependencies
3. Generate task breakdown - Create hierarchical tasks organized by phase/sprint
4. Apply template - Format output using assets/tasks-template.md
5. Review dependencies - Validate task sequencing and parallel opportunities

Input Requirements

Locate specification documents in these common paths:

• PRD: .claude/docs/specs/prd.md or docs/prd.md
• Tech Specs: .claude/docs/specs/tech-specs.md or docs/tech-specs.md
• UI/UX: .claude/docs/specs/ui-ux.md or docs/ui-ux.md

If documents are missing, prompt user for their locations.

Task Generation Process

Step 1: Extract from PRD

Parse the PRD for:

• Core features and user stories → become sprint goals
• MVP requirements → Phase 1/2 priority tasks
• Non-functional requirements → infrastructure/foundation tasks
• Success metrics → verification criteria

Step 2: Extract from Tech Specs

Parse Tech Specs for:

• Architecture decisions → foundation phase setup tasks
• Data models/schemas → type definition tasks (SPRINT-002)
• API contracts → integration tasks
• Technology stack → environment setup tasks (SPRINT-001)
• Performance requirements → optimization tasks

Step 3: Extract from UI/UX Specs

Parse UI/UX for:

• Component specifications → UI development tasks
• User flows → feature integration tasks
• Design system requirements → styling/theming tasks
• Accessibility requirements → QA tasks

Step 4: Organize into Phases

Phase 1: Foundation (Always required)

• SPRINT-001: Environment Setup - tooling, dependencies, project structure
• SPRINT-002: Define Contracts/Types - interfaces, schemas, validation

Phase 2: Build (Feature development)

• Group related features into sprints
• Order by dependency chain
• Mark parallel tasks with [P]

Phase 3: Deployment

• CI/CD pipeline setup
• Environment configuration
• Monitoring and observability

Task ID Convention

Use sequential numbering: T001, T002, T003...

Alternative: Decade-based per sprint: SPRINT-001: T001-T009, SPRINT-002: T010-T019

Choose one approach consistently.

Parallel Task Rules

Mark tasks with [P] when:

• Task has NO dependencies on other tasks in same sprint
• Task can start immediately when sprint begins
• Multiple [P] tasks can run simultaneously

Example:

- T001 [P]: Initialize project structure
- T002 [P]: Configure linting
- T003: Install dependencies (depends on T001)

Acceptance Criteria Guidelines

Each sprint must include:

1. Acceptance Criteria - Checklist of completion requirements
2. Verification - Concrete steps to validate completion

Pattern:

**Acceptance Criteria:**
- [ ] [Specific measurable outcome]
- [ ] [Tests pass / coverage requirement]
- [ ] [Integration verified]

**Verification:**
- Run [specific command] and confirm [expected result]
- Demo [specific feature] to verify [behavior]

Output Format

Use the template at assets/tasks-template.md for consistent formatting.

Key sections:

• Overview with project metadata
• Roadmap organized by Phase → Sprint → Tasks
• Dependencies and Execution Order
• Parallel Opportunities
• Risks & Blockers

Task Estimation Heuristics

When specs don't provide estimates:

• Type definitions: 1-2 tasks per domain
• API endpoints: 1 task per endpoint + integration
• UI components: 1 task per major component
• Environment setup: 3-5 tasks standard
• Testing: ~30% of implementation tasks

References

• For detailed task analysis patterns: see references/task-analysis.md
• For output template: see assets/tasks-template.md