Plan Backlog - Sprint Planning for Parallel Work

Analyze, organize, and prioritize beads issues to enable efficient parallel execution with multiple Claude workers.

Quick Start

/conductor:plan-backlog

Workflow Overview

Phase	Purpose	Reference
1. Analyze	Understand current state	references/plan-backlog/analysis-phases.md
2. Parallelize	Find concurrent opportunities	references/plan-backlog/analysis-phases.md
3. Break Down	Decompose epics	references/plan-backlog/breakdown-prioritize.md
4. Prioritize	Optimize order	references/plan-backlog/breakdown-prioritize.md
5. Plan	Generate sprint waves	references/plan-backlog/sprint-planning.md
6. AI Analysis	Optional Codex insights	references/plan-backlog/sprint-planning.md

---

Quick Analysis

# Current state
bd stats
bd ready
bd blocked

# Count by status
echo "Open: $(bd list --status=open --json | jq 'length')"
echo "Ready: $(bd ready --json | jq 'length')"
echo "Blocked: $(bd blocked --json | jq 'length')"

# High-impact blockers (prioritize these)
bd list --all --json | jq -r '.[] | select(.blocks | length > 0) | "\(.id): blocks \(.blocks | length)"'

# Epic analysis - list children of an epic
bd list --filter-parent <epic-id>        # All issues under epic
bd ready --filter-parent <epic-id>       # Ready issues in epic
bd blocked --filter-parent <epic-id>     # Blocked issues in epic

---

Parallelization Quick Check

# Independent issues - can run in parallel NOW
bd ready --json | jq -r '.[] | "\(.id): [\(.priority)] \(.title)"'

Group by area:

Area	Indicators
Frontend	UI, component, modal, style
Backend	API, server, endpoint
Terminal	xterm, pty, tmux
MCP	tool, mcp, browser

---

Generate Sprint Plan

1. Determine Workers

Ask: How many parallel workers? (2-5, recommend 3)

2. Match Issues to Skills

Workers receive skill hints in prompts based on issue keywords:

Area Keywords	Skill Hint	Purpose
terminal, xterm, pty, resize	/xterm-js:xterm-js	Terminal rendering, WebSocket
UI, component, modal, dashboard	/ui-styling:ui-styling	shadcn/ui, Tailwind patterns
backend, api, server, endpoint	/backend-development:backend-development	APIs, databases
mcp, browser, screenshot	/conductor:tabz-mcp	Browser automation

3. Build Waves

• Wave 1: All bd ready issues (no blockers)
• Wave 2: Issues unblocked after Wave 1
• Wave 3: Issues unblocked after Wave 2

Per-epic wave planning:

# Plan waves for a specific epic
bd ready --filter-parent <epic-id>        # Wave 1 for this epic
bd blocked --filter-parent <epic-id>      # Future waves (resolve deps first)

# Example: Focus a swarm on one epic
bd ready --filter-parent TabzChrome-xyz --json | jq -r '.[].id' | xargs /conductor:bd-swarm

4. Output Format

## Wave 1 (Start Now)
| Issue | Type | Priority | Skill Hint |
|-------|------|----------|------------|
| xxx | feature | P1 | /ui-styling:ui-styling |
| yyy | bug | P2 | /xterm-js:xterm-js |

**Next steps:**
Spawn workers: `/conductor:bd-swarm xxx yyy`

---

Epic Breakdown (Interactive)

# Find epics
bd list --all --json | jq -r '.[] | select(.type == "epic") | .id'

# Analyze epic children (use --filter-parent)
bd list --filter-parent <epic-id>         # All children of epic
bd ready --filter-parent <epic-id>        # Ready children (can start now)
bd blocked --filter-parent <epic-id>      # Blocked children (need deps resolved)

# Count children by status
echo "Total: $(bd list --filter-parent <epic-id> --json | jq 'length')"
echo "Ready: $(bd ready --filter-parent <epic-id> --json | jq 'length')"

# Create subtasks with deps
bd create --title="Design API" --type=task --priority=2
bd create --title="Implement backend" --type=task --priority=2
bd dep add impl-id design-id

---

Priority Adjustments

Criterion	Action
Blocks 3+ issues	Raise to P1
Quick win	Raise to P2
No dependents	Lower to P3
User-facing bug	Raise to P1

bd update <id> --priority 1

---

AI-Assisted (Optional)

For large backlogs (10+ issues), use Codex for analysis:

# Analyze dependency graph
mcp-cli call codex/codex "$(bd list --all --json | jq -Rs '{prompt: "Analyze for parallelization: " + .}')"

See: references/plan-backlog/sprint-planning.md for full Codex prompts.

---

Reference Files

File	Content
analysis-phases.md	Phase 1-2: State analysis, parallelization
breakdown-prioritize.md	Phase 3-4: Epic breakdown, priority tuning
sprint-planning.md	Phase 5-6: Wave generation, AI analysis

---

Notes

• Re-run after each wave completes to re-plan
• Use bd comments <id> add "Progress: ..." for status tracking
• Consider /wipe between major planning sessions
• For complex epics, use ultrathink prefix before breakdown

Execute this workflow now.