scale · 11

Multi-agent: orchestrate & campaign

Add a fresh reviewer, split a goal into merged child runs, or fan one goal out to many gated orchestrators and compose the results.

maps to as-built §30§36

the same gate, lifted up a level · one agent, then many
Add a reviewer, split a goal, or run a campaign Three escalating patterns, each reusing the same gated run. Review runs a coder, then launches a fresh reviewer as an extend of the coder's run, gates it, and produces one result. Full-plan splits a goal into several independent child runs that run in parallel, each gated on its own, then merges them with DAG-aware conflict repair into one result. Campaign lifts that one level higher: it splits a goal into several full-plan orchestrations, each gated, then meta-merges their results, with a hard depth cap of two. REVIEW · a fresh set of eyes coder run cli A · gated extend reviewer cli B · fresh eyes one result ✓ FULL-PLAN · split, run in parallel, merge one goal 2 to 6 tasks child run 1 · gated on its own ✓ child run 2 · gated on its own ✓ child run 3 · gated on its own ✓ merge dependency-aware + repair one result ✓ CAMPAIGN · split into whole orchestrations · depth cap 2 one goal N sub-goals full-plan orchestration 1 ✓ full-plan orchestration 2 ✓ full-plan orchestration 3 ✓ meta-merge tree budget capped one result ✓
Nothing about the gate changes: it just runs in more places. review → full-plan → campaign is the same primitive, fanned out wider.

Once a single gated run works, you can fan it out. The gate never changes: it just runs in more places. There are three escalating patterns.

review: a fresh set of eyes

orchestrate review runs a coder, then launches a reviewer as an extend of the coder's run so lineage is preserved. Point each role at a different CLI (say a Claude coder and a Codex reviewer) and the reviewer verifies independently before anything is accepted.

full-plan: split, run in parallel, merge

orchestrate full-plan asks a planner to split a goal into two to six tasks, runs them as independent child runs (each gated on its own), then merges them. The merge is dependency-aware: if two tasks change the same file, the task that builds on the other wins, and genuine conflicts are sent to a repair provider rather than dumped on you.

campaign: orchestrations of orchestrations

campaign lifts that one level higher: it splits a goal into several full full-plan orchestrations and meta-merges their results. This is the harness of harnesses at the top, a coordinator of coordinators, with a hard depth cap of two and a tree-wide budget so it can't run away.

source