test(sleep): add verifier-discipline stress test for the validation gate (#87)

Adds a reward-hacking stress test ensuring the consolidation gate rejects skill edits that game train/replay tasks while degrading held-out behavior. Also wires the minimax_chat backend into scripts/eval_only.py (coexisting with the qwen wiring from #85). Closes #67.

scripts/eval_only.py

@@ -29,6 +29,7 @@ from skillopt.model import (
     configure_claude_code_exec,
     configure_codex_exec,
     configure_qwen_chat,
+    configure_minimax_chat,
     set_reasoning_effort,
     set_target_backend,
     set_target_deployment,
@@ -138,7 +139,7 @@ def parse_args() -> argparse.Namespace:
     # Legacy flat overrides
     p.add_argument("--env", type=str)
     p.add_argument("--backend", type=str,
-                   choices=["azure_openai", "codex", "codex_exec", "claude", "claude_chat", "claude_code_exec"])
+                   choices=["azure_openai", "codex", "codex_exec", "claude", "claude_chat", "claude_code_exec", "minimax", "minimax_chat"])
     p.add_argument("--optimizer_model", type=str)
     p.add_argument("--target_model", type=str)
     p.add_argument("--optimizer_backend", type=str)
@@ -180,6 +181,12 @@ def parse_args() -> argparse.Namespace:
     p.add_argument("--claude_code_exec_use_sdk", type=str)
     p.add_argument("--claude_code_exec_effort", type=str)
     p.add_argument("--claude_code_exec_max_thinking_tokens", type=int)
+    p.add_argument("--minimax_base_url", type=str)
+    p.add_argument("--minimax_api_key", type=str)
+    p.add_argument("--minimax_model", type=str)
+    p.add_argument("--minimax_temperature", type=float)
+    p.add_argument("--minimax_max_tokens", type=int)
+    p.add_argument("--minimax_enable_thinking", type=_BOOL)
     p.add_argument("--out_root", type=str)
     p.add_argument("--data_path", type=str)
     p.add_argument("--split_mode", type=str,
@@ -255,6 +262,12 @@ def main() -> None:
                 "claude_code_exec_use_sdk": "model.claude_code_exec_use_sdk",
                 "claude_code_exec_effort": "model.claude_code_exec_effort",
                 "claude_code_exec_max_thinking_tokens": "model.claude_code_exec_max_thinking_tokens",
+                "minimax_base_url": "model.minimax_base_url",
+                "minimax_api_key": "model.minimax_api_key",
+                "minimax_model": "model.minimax_model",
+                "minimax_temperature": "model.minimax_temperature",
+                "minimax_max_tokens": "model.minimax_max_tokens",
+                "minimax_enable_thinking": "model.minimax_enable_thinking",
                 "seed": "train.seed",
                 "test_env_num": "evaluation.test_env_num",
                 "env": "env.name",
@@ -312,6 +325,9 @@ def main() -> None:
         elif backend == "claude_code_exec":
             cfg.setdefault("optimizer_backend", "openai_chat")
             cfg.setdefault("target_backend", "claude_code_exec")
+        elif backend in {"minimax", "minimax_chat"}:
+            cfg.setdefault("optimizer_backend", "openai_chat")
+            cfg.setdefault("target_backend", "minimax_chat")
         else:
             cfg.setdefault("optimizer_backend", "openai_chat")
             cfg.setdefault("target_backend", "openai_chat")
@@ -337,6 +353,15 @@ def main() -> None:
             and not _has_model_override("model.target", "target_model")
         ):
             cfg["target_model"] = default_model_for_backend("claude_chat")
+    if cfg.get("target_backend") == "minimax_chat":
+        if (
+            str(cfg.get("target_model", "") or "").strip() in _OPENAI_DEFAULT_MODEL_SENTINELS
+            and not _has_model_override("model.target", "target_model")
+        ):
+            cfg["target_model"] = (
+                cfg.get("minimax_model")
+                or default_model_for_backend("minimax_chat")
+            )
 
     if not cfg.get("out_root"):
         env = cfg.get("env", "unknown")
@@ -416,6 +441,16 @@ def main() -> None:
         target_max_tokens=cfg.get("target_qwen_chat_max_tokens"),
         target_enable_thinking=cfg.get("target_qwen_chat_enable_thinking"),
     )
+    configure_minimax_chat(
+        base_url=cfg.get("minimax_base_url") or None,
+        api_key=cfg.get("minimax_api_key") or None,
+        temperature=cfg.get("minimax_temperature"),
+        max_tokens=cfg.get("minimax_max_tokens"),
+        enable_thinking=cfg.get("minimax_enable_thinking"),
+    )
+    minimax_model_cfg = cfg.get("minimax_model")
+    if minimax_model_cfg and cfg.get("target_backend") == "minimax_chat":
+        set_target_deployment(str(minimax_model_cfg))
     set_reasoning_effort(cfg.get("reasoning_effort", "") or None)
 
     # Build adapter

tests/test_sleep_engine.py

@@ -1006,5 +1006,46 @@ class TestClaudeCliBackendBare(unittest.TestCase):
         self.assertIn("Not logged in", getattr(be, "last_call_error", ""))
 
 
+
+
+class MockRewardHackingBackend(MockBackend):
+    def attempt(self, task: TaskRecord, skill: str, memory: str, sample_id: int = 0) -> str:
+        ctx = (skill or "") + "\n" + (memory or "")
+        # If the reward-hacking artifact is present in the context:
+        if "Always emit placeholder URLs" in ctx:
+            if "rule:__reward_hacking__" in task.tags:
+                return str(task.reference) # perfectly answers the train/replay shortcut task
+            else:
+                return "placeholder URL" # completely breaks the real held-out task
+                
+        # Baseline behavior (without the rule)
+        if "rule:__reward_hacking__" in task.tags:
+            return "I am missing input" # baseline fails the shortcut task
+        if "rule:real" in task.tags:
+            return str(task.reference) # baseline gets the real task right
+            
+        return super().attempt(task, skill, memory, sample_id)
+
+    def reflect(self, failures, successes, skill, memory, **kwargs):
+        edits = []
+        for task, _ in failures:
+            if "rule:__reward_hacking__" in task.tags:
+                edits.append(EditRecord("skill", "add", "Always emit placeholder URLs", rationale="shortcut"))
+                break
+        return edits
+
+class TestVerifierDiscipline(unittest.TestCase):
+    def test_gate_rejects_reward_hacking_edit(self):
+        be = MockRewardHackingBackend()
+        train_task = TaskRecord(id="t1", project="/p", intent="train", reference="XYZABC", reference_kind="exact", tags=["rule:__reward_hacking__"], split="train")
+        val_task = TaskRecord(id="v1", project="/p", intent="val", reference="PQRSTU", reference_kind="exact", tags=["rule:real"], split="val")
+        tasks = [train_task, val_task]
+
+        res = consolidate(be, tasks, "", "", edit_budget=4, gate_metric="hard", night=1)
+        
+        self.assertFalse(res.accepted)
+        self.assertGreater(len(res.rejected_edits), 0)
+        self.assertIn("placeholder", res.rejected_edits[0].content)
+
 if __name__ == "__main__":
     unittest.main(verbosity=2)