work-breakdown
Break down implementation plans into schedulable bd (beads) issues with proper epics, tasks, dependencies, and estimates.
Installation
Follow the repository instructions for this skill and your agent. Review the files and dependencies before adding it to your workflow.
View source ↗GuidesSkill instructions
Work Breakdown Skill
Break down implementation plans into bd-compatible issues. This skill produces epics, tasks, dependencies, and estimates that can be directly executed via bd commands.
When to Invoke
- User provides an implementation plan and asks to break it down
Workflow
Step 1: Read the Implementation Plan
If the user references a file, read it first. Extract:
- Objective: What is the end goal?
- Phases: Are there distinct implementation phases?
- Deliverables: What outputs are expected?
- Dependencies: What depends on what?
Step 2: Identify Epics (Work Packages)
Group work into 3-7 epics. Common patterns:
- Phase 0: Setup/Infrastructure
- Phase 1-N: Feature implementation phases
- Testing & Quality
- Documentation & Deployment
Step 3: Decompose into Tasks
For each epic, create tasks that are:
- 30-480 minutes (0.5-8 hours) of focused work
- Deliverable-oriented: produces a testable output
- Single-responsibility: one executor can complete it
- Ensure your tasks are aligned with the development guidelines in @docs/MAINTENANCE.md
Step 4: Map Dependencies
Identify blocking relationships:
- Task B cannot start until Task A completes
- Epic 2 depends on Epic 1 completion
- Testing depends on implementation
Step 5: Assign Estimates and Priorities
Use bd-compatible values:
- Priority: 0 (critical) to 4 (backlog)
- Estimate: minutes (60=1hr, 240=4hrs)
- Type: epic, task, feature, bug, chore
Step 6: Present WBS for Approval
Show the user:
- Summary table with all tasks
- Dependency graph (text representation)
- Parallel opportunities
- Total effort estimate
Ask: "Ready to create these issues in beads?"
Step 7: Execute bd Commands
After user approval, execute commands in this order:
# 1. Create epics first (capture IDs from --json output)
bd create "Epic Name" -t epic -p [priority] -e [total_minutes] -d "Description" --json
# 2. Create tasks as children of epics
bd create "Task Name" -t task -p [priority] -e [minutes] --parent [epic_id] -d "Description" --json
# 3. Add dependencies between tasks
bd dep add [dependent_id] [blocker_id]
# 4. Show ready work
bd ready
Output Format
Present the WBS as a markdown table:
## Work Breakdown: [Project Name]
### Summary
- Total: N tasks across M epics
- Effort: X-Y hours
- Critical path: Task1 -> Task2 -> Task3
### Epic 1: [Name] (P1, ~Xhr)
| ID | Task | Type | Pri | Est | Depends | Risk |
|----|------|------|-----|-----|---------|------|
| 1.1 | [Title] | task | P2 | 120m | - | Low |
| 1.2 | [Title] | task | P2 | 180m | 1.1 | Med |
### Epic 2: [Name] (P2, ~Xhr)
...
bd Quick Reference
| Field | Flag | Values |
|---|---|---|
| type | -t |
epic, task, feature, bug, chore |
| priority | -p |
0-4 (0=critical) |
| estimate | -e |
minutes |
| parent | --parent |
epic ID |
| description | -d |
text |
| labels | -l |
comma-separated |
Dependencies: bd dep add [child] [blocker] - child depends on blocker
Key Rules
- Always get approval before creating issues
- Create epics first to get their IDs for --parent
- Use --json flag to capture created issue IDs
- Add dependencies after all tasks exist
-
Run
bd readyat the end to show actionable work
Examples
Read the usage examples in the source repository.
View source ↗Source history
- Indexed on
- 2026-01-20
- Index updated
- 2026-01-26
- Branch
- main
Index and sync dates describe this directory. View the source repository for code history.
GitHub ↗chat Comments (0)
Sign in to join the discussion and leave a comment.
No comments yet. Be the first to share your thoughts!