What problem does it solve? Complex engineering tasks often fail when a single agent tries to do everything serially, or when delegation happens without bounded context, independent review, or verifiable completion. This Skill gives the primary AI a just-in-time orchestration protocol for deciding when and how to delegate, compare structural alternatives, and enforce deterministic execution contracts. ## Core Features & Use Cases - Topology Selection: Choose from direct, probe, fan_out, pipeline, writer_reviewer, audit, or custom DAG topologies based on dependencies, ownership, and useful independence rather than a fixed hierarchy. - Bounded Worker Contracts: Project a least-privilege context envelope for each delegated worker, including objective, scope, selected skill bodies, write ownership, verifier, and stop condition. - Deterministic Loop/Graph Runtime: Declare Loop Contracts with goal, verifier, attempt budget, and escalation boundary, and materialize DAGs that reject cycles and overlapping write ownership before execution. - Use Case: When refactoring a service with an expensive-to-reverse persistence seam, use this Skill to compare genuinely different structural alternatives, fan out independent research workers, and verify the final implementation with a command-bound loop receipt instead of narrative confidence. ## Quick Start Ask the agent to use the architect-orchestrator skill to decompose this task, delegate independent work to bounded workers, and verify completion with a deterministic loop contract.
