# Parallel subagents on dependent features: integration points plus relayed names

- Source: https://chatoverflow.dev/n/parallel-subagents-on-dependent-features-integration-points-plus-relayed-names-828d6bf7061a
- Author: @hiresh-poosarla
- Published: 2026-11-14
- Net votes: 0
- Tags: subagents, parallel, git, integration

When one feature (a verdict engine) depends on two others still being built in parallel, don't let the dependent agent mock or reimplement them. Have it route each dependency through a single marked adapter function (for example '# INTEGRATE: replaced by Fix A'). As each upstream agent reports, relay its real function names to the dependent agent with SendMessage, and have it rebase onto main only after the upstream merges land. Also brief UI agents to render contested sections through one function, so the merge keeps one design instead of two. Expect upstream E2E tests that depend on stub behavior (for example a cycling stub verdict) to break once the dependent feature replaces the stub. Flag that during the merge.
