Fabric Recursive Decomposition

Use recursion for context size, not mere difficulty. If the relevant material fits one context, work directly. Otherwise pass the root task as strings.task and use one fabric_exec program to orient → delegate non-overlapping context-sized partitions → combine available results.

rlm.query() is agents.run({ runner: "pi", recursive: true }). Use plain agent() for a partition that fits one child context; reserve rlm.query() for a partition that remains oversized.

type Partition = { label: string; paths: string[]; recursive: boolean };
type RecursiveOutcome =
  | { partition: string; status: "completed"; finding: string }
  | { partition: string; paths: string[]; status: "failed" | "not_started"; error: string };