orchestrator

To decompose large specifications into atomic tasks and prevent AI hallucinations, use deterministic task decomposition to break down work into traceable chunks. This enforces strict context limits (~500 lines) and task sizes (<100 LOC) with clear acceptance criteria, minimizing misinterpretations.

BDD-to-DDD mapping for AI workflows converts given scenarios into domain-driven components and atomic tasks. It encourages a co-located DDD directory structure per aggregate, ensuring work remains cohesive and explicitly mapped to behavioral specifications.

To establish task dependencies and context limits for a 5,000-line spec, apply deterministic task decomposition targeting Phase 3.5. This outputs a tasks.md file with 8-12 atomic tasks, explicit dependency DAGs, and strict context controls (~500 lines per task).

Yes, the orchestrator skill requires analyst and architect dependencies. It relies on these preceding roles to supply the approved specifications needed to generate atomic tasks, establish dependencies, and enforce quality gates.

The best way to enforce quality gates and max task size in AI workflows is through deterministic task decomposition. This approach enforces strict limits (<100 LOC per task, ~500 lines context) and ensures each task has clear acceptance criteria for reliable implementation.

You should not use deterministic task decomposition for AI workflows when dealing with small, loosely defined specifications that do not require strict context limits or dependency DAGs. It is specifically designed for large specifications needing strict quality gates and traceability.