Summary
The create-beads skill frames the "epic + child tasks" vs "standalone task" choice by size ("Multi-task features" vs "Small features"), but the decision that actually matters for how the beads get worked is independence, not size. The skill never prompts that question.
Where
skills/create-beads/SKILL.md → ## Structure:
**Multi-task features:** Epic + child tasks
...
**Small features:** Standalone tasks
...
Expected vs actual
When you have 2+ candidate items, the load-bearing question is: are they independent — no shared files, no sequential dependency?
- Independent items belong as sibling tasks under an epic.
bd ready then surfaces them all at once, so an executor (human or agent) can pick them up in parallel.
- Coupled items (shared files, or one depends on another's outcome) belong in a single bead with steps — splitting them just manufactures dependency edges you then have to manage.
Today, because the only cue is size, two failure modes are easy to fall into:
- Filing several independent items as one bead with a bullet-list of subtasks —
bd ready never surfaces them as parallelizable, so the parallel-dispatch opportunity is lost at authoring time.
- Filing coupled items as separate beads — creating dependency edges and blocked-bead churn for work that was really one unit.
Proposed direction
One approach: add a short subsection to ## Structure that makes the independence question explicit at creation time, e.g.:
Epic or one bead? When you have 2+ candidate items, ask if they're independent — no shared files, no sequential dependency:
- Independent → sibling tasks under one epic.
bd ready surfaces them in parallel.
- Coupled (shared files, or one needs another's outcome) → a single bead with steps.
Default to one bead when unsure — it's cheaper to split later than to merge two.
Happy to open a PR (I have one drafted) — wording/placement is of course yours to decide.
Context
Surfaced while using beadpowers in a downstream workflow project (loom, public) that adopted exactly this independence heuristic in its own bead-authoring conventions and found the same prompt was missing upstream in create-beads. Tracked downstream as loom-7bn.
Summary
The
create-beadsskill frames the "epic + child tasks" vs "standalone task" choice by size ("Multi-task features" vs "Small features"), but the decision that actually matters for how the beads get worked is independence, not size. The skill never prompts that question.Where
skills/create-beads/SKILL.md→## Structure:Expected vs actual
When you have 2+ candidate items, the load-bearing question is: are they independent — no shared files, no sequential dependency?
bd readythen surfaces them all at once, so an executor (human or agent) can pick them up in parallel.Today, because the only cue is size, two failure modes are easy to fall into:
bd readynever surfaces them as parallelizable, so the parallel-dispatch opportunity is lost at authoring time.Proposed direction
One approach: add a short subsection to
## Structurethat makes the independence question explicit at creation time, e.g.:Happy to open a PR (I have one drafted) — wording/placement is of course yours to decide.
Context
Surfaced while using beadpowers in a downstream workflow project (loom, public) that adopted exactly this independence heuristic in its own bead-authoring conventions and found the same prompt was missing upstream in
create-beads. Tracked downstream asloom-7bn.