| name | plan-feature-roadmap |
| description | Break down large features into phased delivery roadmaps with dependency diagrams. Common user requests: - "Create a roadmap for [feature]" - "Break down [feature] into phases" - "^roadmap" (trigger keyword) |
Create a feature roadmap
When to use
- Large features requiring phased delivery
- Features with multiple interconnected tasks
- Work spanning multiple sprints
- Need to visualize dependencies and sequencing
When NOT to use
- Single straightforward tasks
- Well-defined sequential work without dependencies
- Features already broken down in TDD
Process
When user requests a roadmap, follow these steps sequentially:
- Acknowledge before starting: Now creating a delivery roadmap.
- Context gathering: Read existing PRD/TDD/Discovery docs. Search codebase for relevant patterns.
- Clarifying questions: Identify constraints, capacity, risks. Ask user with recommended solutions.
- Break down work: Identify tasks, assign feature-specific IDs, group into phases, map dependencies.
- Create roadmap: Follow structure and guidelines below.
- Validate dependencies: Ensure no circular dependencies, foundational work precedes dependent work.
- Save: Write to
artefacts/roadmap-<feature-slug>.md.
Roadmap document structure
Generated roadmap files should include:
- Document title and summary — Brief overview of the feature and roadmap purpose
- Dependency diagram — Mermaid graph showing tasks grouped by phase with dependencies
- Work breakdown — Detailed task descriptions organized by phase
- Technical considerations — Critical implementation notes, session bridges, atomic operations
- Related documents — Links to PRD, TDD, Discovery docs (if applicable)
Task identification guidelines
Task ID format:
- Use feature-specific prefixes (AUTH, UI, API, DB, etc.)
- Number sequentially within each prefix (AUTH-1, AUTH-2, etc.)
- Keep prefixes short (2-5 characters)
- Choose meaningful prefixes reflecting the work domain
Task naming:
- Use concise, action-oriented names (e.g., "DB schema update", not "Update the database schema")
- Focus on deliverable, not implementation details
- Each task should be independently testable/verifiable
Dependency diagram guidelines
Structure rules:
- Group tasks into
subgraphblocks representing phases - Label subgraphs as "Phase 1: Foundation", "Phase 2: Logic", etc.
Best practices:
- Keep phases focused (3-5 tasks per phase ideal)
Work breakdown guidelines
Format:
### Phase N: Phase Name
* **TASK-ID: Task Title**
* Bullet point describing what needs to be done
* Additional implementation notes
* Technical constraints or requirements
Writing style:
- Use imperative verbs (Add, Implement, Create, Update, Refactor)
- Include specific details (file names, function names, model fields)
- Mention tools/libraries when relevant
- Note compatibility requirements
- Call out transaction boundaries or atomicity needs
Level of detail:
- Enough for a developer to understand scope
- Not full implementation pseudocode (that's TDD territory)
- Focus on what and why, not detailed how
Phasing strategy
Principles:
- Foundation first - enable downstream work
- Logic second - core functionality
- Migration third - data handling
- Interface last - user experience
Considerations:
- Can early phases be deployed without breaking production?
- Which tasks enable the most parallel work downstream?
- What's the minimum viable Phase 1 that unblocks everything else?
Validation checklist
Before finalizing:
- No circular dependencies, all dependencies point forward
- All PRD requirements covered
- Each phase independently deployable
Example roadmap
# Roadmap: User authentication
## Dependency diagram
```mermaid
graph TD
subgraph P1 [Phase 1: Foundation]
AUTH1["AUTH-1: DB Schema"]
end
subgraph P2 [Phase 2: Logic]
AUTH2["AUTH-2: Registration"]
end
subgraph P3 [Phase 3: Interface]
AUTH3["AUTH-3: Login Pages"]
end
AUTH1 --> AUTH2
AUTH2 --> AUTH3
```
## Work breakdown
### Phase 1: Foundation
* **AUTH-1: DB Schema**
* Add `email`, `passwordHash` to User model
* Generate migration with Prisma
### Phase 2: Logic
* **AUTH-2: Registration**
* Server Action for user creation
* Validate email uniqueness, password complexity
### Phase 3: Interface
* **AUTH-3: Login Pages**
* Build `/signup` and `/login` routes
* Form validation and error states
## Technical considerations
- **Compatibility:** Ensure bcrypt works with `nodejs_compat` mode