create-beads

name	create-beads
description	Use when an implementation plan exists and needs to be broken into parallelizable work units with dependencies

Create Beads

Convert implementation plan into beads with dependencies so workers can claim and execute them.

Prerequisites

• Implementation plan exists at docs/plans/*-implementation.md
• bd CLI is available (check with which bd)
• If bd not initialized: bd init --prefix <project-name> --quiet

Process

1. Read the Implementation Plan

Find and read the most recent implementation plan:

ls -t docs/plans/*-implementation.md | head -1

Identify each ### Task N: section as a distinct unit of work.

2. Assess Scaffold Opportunities

Before creating task beads, analyze whether a foundational structure would unlock parallelism.

Scaffolds are:

• Module files with type definitions and function signatures (empty implementations)
• Directory structure and basic interfaces
• Test file skeletons with pending/skip markers
• Data schemas/structs that multiple tasks will use

Benefits:

• Workers can implement different functions independently once scaffold exists
• Easier merges (workers fill in pre-defined slots)
• Guidance (skeleton shows exactly where code goes)

Create scaffold bead if: Multiple tasks modify the same module, or tasks create interdependent types.

3. Assess Dependencies

Analyze the plan to determine task dependencies:

• File references: If Task 3 modifies a file Task 2 creates, Task 3 depends on Task 2
• Logical ordering: Tests for a feature depend on the feature existing
• Parallelization: Independent components can run simultaneously

Default to sequential if uncertain.

4. Create Beads

For scaffold (if identified):

bd create "Create module skeletons for <feature>" -d "Set up file structure and interfaces" -t task -p 1
# Note the returned bead ID (e.g., proj-abc1)

For each task:

bd create "Task N: <title from plan>" -d "<brief description>" -t task -p 2
# Note each returned bead ID

5. Wire Dependencies

# If scaffold exists, it blocks all first-level tasks
bd dep add <task1-id> <scaffold-id>
bd dep add <task2-id> <scaffold-id>

# Tasks block their dependents
bd dep add <task3-id> <task2-id>

6. Report the Dependency Graph

Output the created structure:

Created N beads from implementation plan:

Scaffold: proj-a1b2 - Create module skeletons
  |-- Task 1: proj-c3d4 - Implement user module
  |-- Task 2: proj-e5f6 - Implement auth module (parallel with Task 1)
      |-- Task 3: proj-g7h8 - Wire integration
          |-- Task 4: proj-i9j0 - Final verification

Ready to start: proj-a1b2 (Scaffold)

Verify with: bd ready

Error Handling

bd not initialized: Run bd init --prefix <project-name> --quiet first. Detect project name from directory.

bd routes to different database: If bd --verbose shows "Routing to target repo: ", the CLI has multi-repo routing configured. Either:

• Use --db .beads/beads.db to force local database
• Check that the target database has issue_prefix configured
• Or work with the routing as intended (beads go to planning repo)

Beads already exist for this plan: Check bd list for matching titles. Warn and ask whether to skip, update, or abort.

Plan has no clear task structure: Fall back to single bead for entire plan. Flag: "Could not identify distinct tasks."

Dependency analysis uncertain: Default to sequential. Add note: "Dependencies assumed sequential - review for parallelization."

Output

• Beads appear in bd ready for workers to claim
• BeadsWatcher detects .beads/issues.jsonl changes
• BeadsCoordinator spawns workers for ready beads