feat(libsy): add plan-execute routing strategy

README.md

@@ -327,6 +327,7 @@ Most use an LLM as a judge. All of them pick between an **efficient** model and
 |---|---|---|---|
 | **[Capability](docs/routing_algorithms/llm_classifier_routing.md)** | The first request is judged by an LLM. | `llm_classifier` | 71.2% at $79.32 |
 | **[Stage](docs/routing_algorithms/stage_router_routing.md)** | Tool responses are judged by pattern matching or an LLM. | `stage_router` | 72.7% at $68.19 |
+| **[Plan/Execute](docs/routing_algorithms/plan_execute_routing.md)** | A capable model plans, then an efficient model executes after the first edit. | `plan_execute` | not yet benchmarked |
 | **[Capability + Stage](docs/routing_algorithms/composite_routing.md)** | Combines the two above. | `composite` | not yet benchmarked |
 | **[Escalation](docs/routing_algorithms/escalation_router_routing.md)** | Starts efficient. Responses are judged by an LLM for issues, then escalated. | `llm_classifier` + `mode = "escalation"` | 75.7% at $85.00 |
 | **[Advisor Gate](docs/routing_algorithms/advisor_gate_routing.md)** | One model serves every turn; a stronger advisor approves its plans and "done" claims, or sends it back. | `advisor` | lifts a weak executor 43.8% → 54.7% |

benchmark/README.md

@@ -193,6 +193,22 @@ benchmark/server-configs/tb-lite-single-gpt-5-5.toml
 benchmark/server-configs/tb-lite-single-opus-4-7.toml
 ```
 
+To benchmark the plan/execute boundary, use
+`benchmark/server-configs/tb-lite-plan-execute-opus-kimi.toml`. It starts each task on Opus and
+hands the preserved trajectory to Kimi after the first edit or write tool call:
+
+```bash
+bash benchmark/run-baseline.sh \
+  --harbor-path benchmark/datasets/openthoughts-tblite-closed-book \
+  --server-config benchmark/server-configs/tb-lite-plan-execute-opus-kimi.toml \
+  --model switchyard \
+  --agent codex \
+  --reasoning-effort xhigh \
+  --n-tasks 1 \
+  --n-concurrent 1 \
+  --max-retries 0
+```
+
 By default, the runner starts in the background and prints the PID, log path, and kill command.
 
 ## Book Modes

benchmark/server-configs/tb-lite-plan-execute-opus-kimi.toml

@@ -0,0 +1,26 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Plan/execute configuration for Harbor Terminal-Bench Lite.
+# Harbor selects this route with --model switchyard.
+
+schema_version = 1
+
+[llm_clients.openrouter]
+format = "openai_chat"
+base_url = "https://openrouter.ai/api/v1"
+api_key_env = "OPENROUTER_API_KEY"
+
+[targets.capable]
+id = "anthropic/claude-opus-4.7"
+llm_client = "openrouter"
+
+[targets.efficient]
+id = "moonshotai/kimi-k2.7-code"
+llm_client = "openrouter"
+
+[routes.switchyard]
+id = "switchyard"
+type = "plan_execute"
+capable_target = "capable"
+efficient_target = "efficient"

crates/libsy/src/algorithms.rs

@@ -13,6 +13,7 @@ pub mod fall_through;
 pub mod llm_class;
 pub mod noop;
 pub mod passthrough;
+pub mod plan_execute;
 pub mod rand;
 pub mod stage;
 pub mod subagent;