Add SWE debug environment

assets/lab/environments/AGENTS.md

@@ -911,6 +911,7 @@ Newer and more experimental environment classes include:
     )
     ```
 - **V1 `vf.Env` / `vf.Taskset` / `vf.Harness`** — preferred taskset/harness pattern for composing task data and program execution without subclassing. Use this for new environments that need reusable tasksets, reusable harnesses, config-driven metrics, rewards, toolsets, users, endpoint interception, or sandboxed Python/command programs. `vf.Taskset` owns train/eval rows, prompt shaping, setup/update/reward hooks, and toolsets. `vf.Harness` owns the framework program, endpoint proxy, model controls, sandbox options, and runtime hooks. `vf.Env` wires them into the standard evaluation and training surface.
+- **`SWEDebugEnv`** — no-agent debugger for SWE-style `SandboxTaskSet` instances. It creates the task sandbox, optionally runs `taskset.setup(state)`, performs one debug step (`none`, `gold_patch`, `command`, or `script`), and optionally runs the task tests and scorer. It records setup, sandbox creation, gold patch, debug command, and test timings in state for validation and timing investigations.
 - **`HarborEnv`** — loads Harbor-format agent benchmark tasks
 - **`RLMEnv`** — implements [Recursive Language Models](https://alexzhang13.github.io/blog/2025/rlm/) for unbounded context processing via REPL-based decomposition and recursive sub-LLM calls
 - **`OpenCodeEnv`** — runs [OpenCode](https://opencode.ai) CLI agents inside sandboxes with API call interception

docs/environments.md

@@ -905,6 +905,7 @@ Newer and more experimental environment classes include:
     )
     ```
 - **V1 `vf.Env` / `vf.Taskset` / `vf.Harness`** — preferred taskset/harness pattern for composing task data and program execution without subclassing. Use this for new environments that need reusable tasksets, reusable harnesses, config-driven metrics, rewards, toolsets, users, endpoint interception, or sandboxed Python/command programs. `vf.Taskset` owns train/eval rows, prompt shaping, setup/update/reward hooks, and toolsets. `vf.Harness` owns the framework program, endpoint proxy, model controls, sandbox options, and runtime hooks. `vf.Env` wires them into the standard evaluation and training surface.
+- **`SWEDebugEnv`** — no-agent debugger for SWE-style `SandboxTaskSet` instances. It creates the task sandbox, optionally runs `taskset.setup(state)`, performs one debug step (`none`, `gold_patch`, `command`, or `script`), and optionally runs the task tests and scorer. It records setup, sandbox creation, gold patch, debug command, and test timings in state for validation and timing investigations.
 - **`HarborEnv`** — loads Harbor-format agent benchmark tasks
 - **`RLMEnv`** — implements [Recursive Language Models](https://alexzhang13.github.io/blog/2025/rlm/) for unbounded context processing via REPL-based decomposition and recursive sub-LLM calls
 - **`OpenCodeEnv`** — runs [OpenCode](https://opencode.ai) CLI agents inside sandboxes with API call interception

docs/reference.md

@@ -516,6 +516,35 @@ class OpenEnvEnv(MultiTurnEnv):
 
 OpenEnv integration that runs OpenEnv projects in Prime Sandboxes using a prebuilt image manifest (`.build.json`), supports both gym and MCP contracts, and requires a `prompt_renderer` to convert observations into chat messages.
 
+#### SWEDebugEnv
+
+```python
+class SWEDebugEnv(SandboxMixin, MultiTurnEnv):
+    def __init__(
+        self,
+        taskset: SandboxTaskSet,
+        dataset: Any = None,
+        *,
+        run_setup: bool = True,
+        debug_step: Literal["none", "gold_patch", "command", "script"] = "gold_patch",
+        run_tests: bool = True,
+        debug_command: str | None = None,
+        debug_script: str | None = None,
+        debug_script_path: str | None = None,
+        debug_timeout: int | None = None,
+        test_timeout: int = 900,
+        cpu_cores: int | None = None,
+        memory_gb: int | None = None,
+        disk_size_gb: int | None = None,
+        labels: list[str] | None = None,
+        timeout_seconds: float = 1800.0,
+        output_tail_chars: int = 2000,
+        **sandbox_kwargs,
+    ): ...
+```
+
+No-agent debugger for SWE-style `SandboxTaskSet` instances. It creates the task sandbox, optionally runs task setup, runs one debug step (`none`, `gold_patch`, `command`, or `script`), and optionally runs tests and scores the result.
+
 #### EnvGroup
 
 ```python

environments/AGENTS.md

@@ -911,6 +911,7 @@ Newer and more experimental environment classes include:
     )
     ```
 - **V1 `vf.Env` / `vf.Taskset` / `vf.Harness`** — preferred taskset/harness pattern for composing task data and program execution without subclassing. Use this for new environments that need reusable tasksets, reusable harnesses, config-driven metrics, rewards, toolsets, users, endpoint interception, or sandboxed Python/command programs. `vf.Taskset` owns train/eval rows, prompt shaping, setup/update/reward hooks, and toolsets. `vf.Harness` owns the framework program, endpoint proxy, model controls, sandbox options, and runtime hooks. `vf.Env` wires them into the standard evaluation and training surface.
+- **`SWEDebugEnv`** — no-agent debugger for SWE-style `SandboxTaskSet` instances. It creates the task sandbox, optionally runs `taskset.setup(state)`, performs one debug step (`none`, `gold_patch`, `command`, or `script`), and optionally runs the task tests and scorer. It records setup, sandbox creation, gold patch, debug command, and test timings in state for validation and timing investigations.
 - **`HarborEnv`** — loads Harbor-format agent benchmark tasks
 - **`RLMEnv`** — implements [Recursive Language Models](https://alexzhang13.github.io/blog/2025/rlm/) for unbounded context processing via REPL-based decomposition and recursive sub-LLM calls
 - **`OpenCodeEnv`** — runs [OpenCode](https://opencode.ai) CLI agents inside sandboxes with API call interception

verifiers/envs/experimental/__init__.py

@@ -8,6 +8,7 @@
     "TaskSet",
     "Harness",
     "ComposableEnv",
+    "SWEDebugEnv",
 ]
 
 
@@ -19,6 +20,7 @@ def __getattr__(name: str):
         "TaskSet": "verifiers.envs.experimental.composable:TaskSet",
         "Harness": "verifiers.envs.experimental.composable:Harness",
         "ComposableEnv": "verifiers.envs.experimental.composable:ComposableEnv",
+        "SWEDebugEnv": "verifiers.envs.experimental.composable:SWEDebugEnv",
     }
     if name in _lazy:
         import importlib

verifiers/envs/experimental/composable/__init__.py

@@ -7,6 +7,7 @@
 )
 from verifiers.envs.experimental.composable.harness import Harness
 from verifiers.envs.experimental.composable.composable_env import ComposableEnv
+from verifiers.envs.experimental.composable.swe_debug_env import SWEDebugEnv
 
 __all__ = [
     "SandboxSpec",
@@ -15,5 +16,6 @@
     "SandboxTaskSet",
     "Harness",
     "ComposableEnv",
+    "SWEDebugEnv",
     "discover_sibling_dir",
 ]