perf(dev-auto): make review-layer fan-out atomic

[alexeyv]

What

Tighten the dev-auto review step so reviewer subagents launch as one atomic fan-out.

step-04-review.md previously said only to "execute all remaining layers in parallel wherever their execution methods allow" — wording that permits an orchestrator to spawn one reviewer, react to its output, then spawn the next.

Why

Observed in a real run: four reviewers (adversarial, edge-case, verification-gap, intent-alignment) launched ~2m14s apart end to end because the parent reacted to earlier output before spawning the rest. Pure wall-clock waste with no quality benefit.

Change

One added sentence:

Spawn every reviewer subagent before reading or reacting to any of their output; begin collection and triage only once all are launched.

No change to reasoning structure or review discipline — only the launch ordering.

🤖 Generated with Claude Code

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 8960acd2-5298-4595-b5db-dc97104f261f

📥 Commits

Reviewing files that changed from the base of the PR and between 52e4c63 and b3168e4.

📒 Files selected for processing (1)

• src/bmm-skills/4-implementation/bmad-dev-auto/step-04-review.md

---

📝 Walkthrough

Walkthrough

This change updates documentation in step-04-review.md to add an explicit ordering constraint: all reviewer subagents must be spawned in parallel before reading or reacting to any reviewer output, replacing the prior instruction that lacked this constraint.

Changes

Review Step Instruction Update

Layer / File(s)	Summary
Spawn-before-read ordering constraint src/bmm-skills/4-implementation/bmad-dev-auto/step-04-review.md	Adds a requirement to launch all reviewer subagents in parallel before reading or reacting to any reviewer output, replacing the prior instruction that only required running layers in parallel.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

• bmad-code-org/BMAD-METHOD#2526: Prior update to the same step-04-review.md parallel subagent launch flow.
• bmad-code-org/BMAD-METHOD#2550: Adds a matching "spawn all reviewer subagents in parallel before reading outputs" ordering constraint.
• bmad-code-org/BMAD-METHOD#2069: Similar subagent orchestration ordering change for a different step file.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the main change: making review-layer fan-out atomic.
Description check	✅ Passed	The description matches the changeset and explains the atomic reviewer launch ordering update.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch chore/dev-auto-atomic-reviewer-fanout

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[augmentcode]

🤖 Augment PR Summary

Summary: Tightens the bmad-dev-auto Step 4 review instructions to enforce an atomic fan-out when launching reviewer subagents.
Why: Avoids sequential reviewer spawning driven by early outputs, reducing wall-clock time without changing the review reasoning/triage structure.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. No suggestions at this time.

Comment augment review to trigger a new review at any time.