Tasks
Task = Individual Work Units
Tasks represent discrete, implementable units of work with clear completion criteria and specific collaboration patterns. Each task includes comprehensive specifications for both human understanding and AI execution.
Task naming convention
Tasks follow a structured naming pattern that indicates their relationship to epics and execution sequence.
Format: [Task X.Y] Descriptive name
X = Epic identifier (1, 2A, 11, etc.)
Y = Sequential task number within epic (1, 2, 3...)
Examples:
[Task 1.1] Define user authentication interface
[Task 2A.3] Implement payment validation logic
[Task 11.2] Configure CI/CD pipeline
Task classification
Every task receives a classification that determines the collaboration pattern between humans and AI.
AI-Ready Tasks
AI-Ready Tasks can be completed independently by AI with minimal human oversight. They are usually single file output, clear specifications, established patterns. They usually show a success rate of 90%+ first-pass approval.
Human Role: Define requirements, review output
AI Role: Full implementation, testing, documentation
Examples:
API endpoint implementations with OpenAPI specifications, database entities following established patterns, React components matching design system specifications, test suite generation for defined business rules, configuration and deployment scripts
AI-Assisted Tasks
This type of tasks require collaborative development between humans and AI during the coding activities. They are very likely multi-component changes, business logic or/and system integrations and they shows regularly 70%+ success rate with minor modifications.
Human Role: Guide approach, validate business logic
AI Role: Generate scaffolding, handle boilerplate, create tests
Examples
Service layer implementations with complex business rules, data migration scripts with validation logic, third-party API integrations with error handling, complex form workflows with validation logic
Hybrid Tasks
Hybrid Tasks are defined as shared ownership with continuous collaboration beween an AI coding assistant and human programmer. Their main characteristic is about strategic implementations and they require business domain expertise to handle the edge cases.
Human Role: Architectural decisions, critical logic
AI Role: Support implementation, generate utilities
Examples
Authentication and authorization systems, payment processing with compliance requirements, workflow orchestration engines, data privacy and GDPR compliance features
Human-Only Tasks
Human-only Tasks require human expertise for critical decisions on complex architectures, compliance, esecurity, etc. Being named Human-Only doesn't mean AI coding assistant isn't involed, but its role will be less driving comparing with the other types.
Human Role: Full design and implementation leadership
AI Role: Research, boilerplate, documentation support
Examples
System architecture and technology choices, security threat modeling and compliance strategy, business process design and optimization, technical debt remediation planning
Task components
Every task must include the following important components. The presence, and good quality of these elements will determine the success of AI in fulfilling its job, and ultimately the qulity of work.
Epic reference
Part of
Epic [X]: [Epic Name]
Dependencies
"Depends on: #X, #Y" or "No dependencies"
Used for wave assignment calculation
Drives cascade analysis when task completes
Wave assignment
Wave: wave-001
Required field for all tasks
Follows dependency analysis algorithm
Supports Epic Integrity Principle
Testing requirements
Test File:
[path to test file]Coverage Target:
[X]%(varies by AI classification)Test Type:
[Unit/Integration/E2E/Type/Manual]Test Scenarios:
Happy path:
[main success scenario]Error case:
[primary failure scenario]Edge case:
[boundary condition]
AI context package
Task Documentation
Here is a general template for the specs compliance for tasks:
Task Definition:
id: "[Task X.Y] Task Name"
classification: "ai-ready | ai-assisted | hybrid | human-only"
confidence_score: 1-10
wave: "wave-001"
epic_reference: "Part of Epic [X]: [Epic Name]"
dependencies: "Depends on: #X, #Y" or "No dependencies"
Acceptance Criteria:
- "Specific, measurable deliverable"
- "Test requirements and coverage targets"
- "Integration points and validation steps"
Testing Requirements:
test_file: "[path to test file]"
coverage_target: "[X]%"
test_scenarios:
- "Happy path"
- "Error case"
- "Edge case"
AI Context Package:
files_to_analyze:
- "src/interfaces/auth.ts"
- "docs/patterns.md"
patterns_to_follow:
- "Repository pattern"
- "Error handling"
constraints:
- "Performance <200ms"
- "Security requirements"
quality_requirements:
- "Test coverage ≥90%"
- "Integration tests"Task sizing
Here are some general guidelines and empirical observations, not required rules. In Hydro, and AI-human hybrid work generally, execution time is less important than the quality of task definitions and and dependency identification pre-work.
AI-Ready Tasks: 1-2 hours (single component scope)
AI-Assisted Tasks: 3-5 hours (bounded complexity)
Hybrid Tasks: 1-2 days (strategic implementations)
Human-Only Tasks: 1-2 days maximum (split if larger)
When to split tasks?
Touches more than 3 files (except human-only integration tasks)
Has more than 5 acceptance criteria
Requires totally different skill sets, security or architecture scope
Definition of Done (standard for all tasks)
The DoD establishes the standard completion criteria that every task must satisfy before transitioning to "Done" status. This checklist serves as the final validation gate in the workflow.
Last updated