What problem does it solve?
Prevents wasted tokens and reviewer fatigue by deciding whether a multi-step AI coding task should run in parallel or stay sequential, based on independence, blast radius, reviewability, and budget.
Core Features & Use Cases
- Decision gate for parallelization: Outputs SEQUENTIAL or PARALLEL-CANDIDATE before subagent execution begins.
- Independence + blast-radius validation: Avoids race conditions and conflicting edits by checking shared state and failure isolation.
- Briefing template for parallel subagents: Ensures each dispatched unit receives a self-contained prompt and a strict deliverable contract.
- Coordination and budgeting guardrails: Caps concurrency (≤5) and recommends batching when subagent counts get large.
- Bilingual guidance (VI/EN): Supports both Vietnamese and English users with consistent operational instructions.
- Cross-track applicability: Applies to angular-portal, nestjs, nextjs, and integrates with sdcorejs-subagent-driven-dev when parallel is chosen.
Quick Start
Invoke sdcorejs-parallel-dispatch when preparing to dispatch 3+ independent units of work (e.g., generating multiple entity screen groups or auditing multiple stacks) so it can choose PARALLEL-CANDIDATE and brief sdcorejs-subagent-driven-dev.
