executing-plans

superbeads:executing-plans

Announce: Using superbeads:executing-plans.

Execution is driven entirely from beads. There is no markdown plan file — the plan lives as child tasks under a beads epic, with bite-sized implementation steps stored in each task's design field. Do not restate the field mapping here — see skills/using-superbeads/SKILL.md.

The Loop

Step 1 — Identify the epic

The epic id is passed explicitly from the superbeads:writing-plans handoff (e.g., superbeads-9i2). If the id is not already in context, ask the user or run bd list --type=epic to identify it. All bd commands below are scoped to that epic's children.

Step 2 — Pick next task

bd ready --parent <epic-id>

bd ready hides blocked tasks; the list is already dependency-ordered. Pick the first item. If the list is empty, go to Step 6.

Step 3 — Read the task

bd show <task-id>

The design field contains the bite-sized steps (code, tests, commands) for this task. Read it in full before touching code.

GATE: If the task's design field is missing, empty, or contradictory, STOP — do not claim it; surface it to the user (the plan task is underspecified).

Step 4 — Claim and implement

bd update <task-id> --claim

Implement using REQUIRED SUB-SKILL: superpowers:test-driven-development — invoke it for this task. Follow its RED-GREEN-REFACTOR cycle using the steps from the design field as the spec.

Commit after the task passes:

git add <changed-files>
git commit -m "feat(<task-id>): <short description>"

Step 5 — Close and surface next work

bd close <task-id> --suggest-next

--suggest-next surfaces newly-unblocked tasks after closing. Then re-run bd ready --parent <epic-id> — that is the source of truth for what is actually unblocked and ready; do not rely on parsing --suggest-next output to drive the loop. Go to Step 2.

Step 6 — Epic drained

When bd ready --parent <epic-id> returns no tasks, all work under this epic is complete.

REQUIRED SUB-SKILL: Invoke superpowers:finishing-a-development-branch to verify, present merge/PR options, and execute the chosen path.

(Note: superbeads:finishing-a-development-branch is a planned follow-on wrapper — not yet built. Fall back to the superpowers original unchanged.)

When to Stop

Stop and ask for help when:

• A task's design field is missing or contradictory
• A test fails repeatedly and the cause is unclear
• bd ready returns tasks that depend on external work not yet done
• bd close errors (task not in the right state)

Do not guess through blockers.

References

• skills/using-superbeads/SKILL.md — source-of-truth rule and field mapping (task steps live in design)
• superpowers:test-driven-development — required for each task's implementation
• superpowers:finishing-a-development-branch — required after epic is drained