The fastest, cheapest, most efficient open-source agent harness. Run more than 500 agents on a $10 VPS.
Rust agent-orchestrationai-agentsai-assistantdesktopllmlocal-firstmcppersonal-aiprivacyrustsecond-braintauri
test(agent-harness): update serialization helper calls for new signature
Steven Enamakel24e8d36ac900
diff --git a/tests/agent_harness_e2e.rs b/tests/agent_harness_e2e.rsindex 3d182e04c2..dad481e2c2 100644--- a/tests/agent_harness_e2e.rs+++ b/tests/agent_harness_e2e.rs@@ -1895,8 +1895,9 @@ async fn subagent_delegation_happy_path_inner() {// relays in its own reply;// * the user's next message resumes the thread with that relay in context.//-// Deterministic request ordering (`serialize_children_after_parent(2)` holds the-// child's first model call until the parent has answered twice):+// Deterministic request ordering holds the child until the parent has answered+// twice, then holds later parent requests until the child's scripted exchange+// is complete:// request[0] = orchestrator turn 1 → spawn_subagent(task_manager_agent)// request[1] = orchestrator, acknowledges; turn 1 ends (chat_done)// request[2] = task_manager_agent → ask_user_clarification pauses the child@@ -1917,7 +1918,7 @@ fn delegated_clarification_flow() {async fn delegated_clarification_flow_inner() {let _lock = env_lock();- serialize_children_after_parent(2);+ serialize_children_after_parent(2, 2);reset_script(vec![tool_call_completion("spawn_subagent",@@ -3693,7 +3694,7 @@ async fn agent_installs_a_registry_skill_then_runs_it_inner() {// The hand-off is a `spawn_subagent` (asynchronous): the parent's// acknowledgement is answered before the child's first call, and the child's// result is delivered to a later parent turn that loads the skill.- serialize_children_after_parent(2);+ serialize_children_after_parent(2, 3);reset_script(vec![tool_call_completion("spawn_subagent",