Add CybORG arena

.github/mlc_config.json

@@ -27,6 +27,9 @@
     {
       "pattern": "https://huskybench\\.com/.*"
     },
+    {
+      "pattern": "https://corewar\\.co\\.uk"
+    },
     {
       "pattern": "https?://(.*\\.)?twitter\\.com/.*"
     },
@@ -35,6 +38,9 @@
     },
     {
       "pattern": "https://www\\.contributor-covenant\\.org/version/2/1/code_of_conduct\\.html"
+    },
+    {
+      "pattern": "https://join\\.slack\\.com/t/swe-bench/shared_invite/.*"
     }
   ]
 }

README.md

@@ -98,12 +98,16 @@ The winner is the LM agent who wins the most rounds.
 ## 🧩 Available Arenas
 
 CodeClash includes competitive programming games and simulation-backed arenas, including BattleSnake,
-CoreWar, Halite, HuskyBench, RoboCode, RobotRumble, and SCML.
+CoreWar, CybORG, Halite, HuskyBench, RoboCode, RobotRumble, and SCML.
 
 SCML is a supply-chain negotiation arena based on the ANAC Supply Chain Management League OneShot
 track. Agents edit a Python `scml_agent.py` implementation and compete to maximize average profit
 across multiple simulated supply-chain worlds.
 
+CybORG is a simulated cyber-defense arena based on the CAGE Challenge 3 DroneSwarm scenario. Agents
+edit a Python `cyborg_agent.py` implementation and compete to maximize blue-team reward across
+simulated episodes.
+
 ## 🚀 Get Involved
 
 - Check out our [docs](https://docs.codeclash.ai/) for more details on running different arenas, configuring tournaments, etc.

codeclash/arenas/__init__.py

@@ -6,6 +6,7 @@
 from codeclash.arenas.bridge.bridge import BridgeArena
 from codeclash.arenas.chess.chess import ChessArena
 from codeclash.arenas.corewar.corewar import CoreWarArena
+from codeclash.arenas.cyborg.cyborg import CybORGArena
 from codeclash.arenas.dummy.dummy import DummyArena
 from codeclash.arenas.figgie.figgie import FiggieArena
 from codeclash.arenas.gomoku.gomoku import GomokuArena
@@ -25,6 +26,7 @@
     BridgeArena,
     ChessArena,
     CoreWarArena,
+    CybORGArena,
     DummyArena,
     FiggieArena,
     GomokuArena,

codeclash/arenas/cyborg/CybORG.Dockerfile

@@ -0,0 +1,26 @@
+FROM python:3.11-slim-bookworm
+
+ENV DEBIAN_FRONTEND=noninteractive \
+    PYTHONDONTWRITEBYTECODE=1 \
+    PIP_NO_CACHE_DIR=1
+
+RUN apt-get update \
+ && apt-get install -y --no-install-recommends \
+    ca-certificates git build-essential jq \
+ && rm -rf /var/lib/apt/lists/*
+
+RUN python -m pip install --upgrade pip \
+ && git clone https://github.com/cage-challenge/CybORG.git /opt/CybORG \
+ && cd /opt/CybORG \
+ && git checkout a2d03f99e587af153ae0ac50fb94ba6272e4fff2 \
+ && python -m pip install "numpy<1.24" -e /opt/CybORG
+
+WORKDIR /workspace
+
+COPY codeclash/arenas/cyborg/runtime/ /workspace/
+
+RUN git init \
+ && git config user.email "player@codeclash.com" \
+ && git config user.name "Player" \
+ && git add . \
+ && git commit -m "Initial CybORG workspace"

codeclash/arenas/cyborg/__init__.py

@@ -0,0 +1,3 @@
+from codeclash.arenas.cyborg.cyborg import CybORGArena
+
+__all__ = ["CybORGArena"]

codeclash/arenas/cyborg/cyborg.py

@@ -0,0 +1,128 @@
+import json
+import shlex
+import subprocess
+
+from codeclash.agents.player import Player
+from codeclash.arenas.arena import CodeArena, RoundStats
+from codeclash.constants import RESULT_TIE
+from codeclash.utils.environment import assert_zero_exit_code
+
+RESULTS_JSON = "cyborg_results.json"
+
+
+class CybORGArena(CodeArena):
+    name: str = "CybORG"
+    submission: str = "cyborg_agent.py"
+    description: str = """CybORG is a simulated cyber-defense arena based on the CAGE Challenge 3 DroneSwarm scenario.
+
+Your bot is a Python file named `cyborg_agent.py` that defines a class named `MyAgent`.
+`MyAgent` should inherit from a CybORG BaseAgent-compatible class, for example:
+
+    from CybORG.Agents import RandomAgent
+
+    class MyAgent(RandomAgent):
+        ...
+
+Each round evaluates every submitted agent independently on the same seeded DroneSwarm episodes.
+Your agent controls the blue-team drone agents through CybORG's simulated PettingZoo interface.
+The objective is to maximize average episode reward. This arena uses CybORG simulation only and does
+    not run real exploit tools or interact with external networks.
+    """
+    default_args: dict = {
+        "steps_per_episode": 30,
+        "num_drones": 18,
+        "timeout": 240,
+    }
+
+    def _game_arg(self, key: str):
+        return self.game_config.get("args", {}).get(key, self.default_args[key])
+
+    def _episodes_per_round(self) -> int:
+        return int(self.game_config.get("args", {}).get("episodes_per_round", self.game_config["sims_per_round"]))
+
+    def validate_code(self, agent: Player) -> tuple[bool, str | None]:
+        quoted_submission = shlex.quote(self.submission)
+        file_check = agent.environment.execute(f"test -f {quoted_submission} && echo exists")
+        if "exists" not in file_check["output"]:
+            return False, f"Submission file `{self.submission}` not found in the workspace root"
+
+        content = agent.environment.execute(f"cat {quoted_submission}")["output"]
+        if not content.strip():
+            return False, f"`{self.submission}` is empty"
+
+        syntax_check = agent.environment.execute(f"python -m py_compile {quoted_submission}")
+        if syntax_check["returncode"] != 0:
+            return False, f"Python syntax error in `{self.submission}`:\n{syntax_check['output']}"
+
+        import_check = agent.environment.execute(
+            "python - <<'PY'\n"
+            "import importlib.util\n"
+            f"spec = importlib.util.spec_from_file_location('submission_agent', {self.submission!r})\n"
+            "module = importlib.util.module_from_spec(spec)\n"
+            "spec.loader.exec_module(module)\n"
+            "assert hasattr(module, 'MyAgent'), 'MyAgent class not found'\n"
+            "from CybORG.Agents import BaseAgent\n"
+            "assert issubclass(module.MyAgent, BaseAgent), 'MyAgent must inherit from a CybORG BaseAgent class'\n"
+            "PY"
+        )
+        if import_check["returncode"] != 0:
+            return False, f"Could not import `MyAgent` from `{self.submission}`:\n{import_check['output']}"
+
+        return True, None
+
+    def execute_round(self, agents: list[Player]) -> None:
+        agent_args = []
+        for agent in agents:
+            agent_args.extend(["--agent", f"{agent.name}=/{agent.name}/{self.submission}"])
+
+        cmd = [
+            "python",
+            "run_cyborg.py",
+            "--episodes",
+            str(self._episodes_per_round()),
+            "--steps",
+            str(self._game_arg("steps_per_episode")),
+            "--drones",
+            str(self._game_arg("num_drones")),
+            "--output",
+            str(self.log_env / RESULTS_JSON),
+            *agent_args,
+        ]
+        full_cmd = " ".join(shlex.quote(part) for part in cmd)
+        self.logger.info(f"Running game: {full_cmd}")
+        try:
+            response = self.environment.execute(full_cmd, timeout=int(self._game_arg("timeout")))
+        except subprocess.TimeoutExpired as exc:
+            raise RuntimeError("CybORG round timed out") from exc
+        assert_zero_exit_code(response, logger=self.logger)
+
+    def get_results(self, agents: list[Player], round_num: int, stats: RoundStats):
+        result_file = self.log_round(round_num) / RESULTS_JSON
+        if not result_file.exists():
+            self.logger.error(f"Missing result file: {result_file}")
+            stats.winner = RESULT_TIE
+            for agent in agents:
+                stats.scores[agent.name] = 0.0
+                stats.player_stats[agent.name].score = 0.0
+            return
+
+        with open(result_file) as f:
+            result = json.load(f)
+
+        scores = {agent.name: 0.0 for agent in agents}
+        for player, score in result.get("average_scores", {}).items():
+            if player in scores:
+                scores[player] = float(score)
+
+        stats.scores = scores
+        stats.details = result.get("details", [])
+        for player, score in scores.items():
+            stats.player_stats[player].score = score
+
+        if not scores:
+            stats.winner = RESULT_TIE
+            return
+
+        top_score = max(scores.values())
+        winners = [player for player, score in scores.items() if score == top_score]
+        stats.winner = winners[0] if len(winners) == 1 else RESULT_TIE

codeclash/arenas/cyborg/runtime/.gitignore

@@ -0,0 +1,2 @@
+__pycache__/
+*.py[cod]

codeclash/arenas/cyborg/runtime/README.md

@@ -0,0 +1,15 @@
+# CybORG CodeClash Workspace
+
+Edit `cyborg_agent.py`.
+
+Your file must define `MyAgent`, a CybORG `BaseAgent` subclass. A safe starting point is:
+
+```python
+from CybORG.Agents import RandomAgent
+
+
+class MyAgent(RandomAgent):
+    pass
+```
+
+The arena runs simulated CAGE Challenge 3 DroneSwarm episodes and scores agents by average reward.

codeclash/arenas/cyborg/runtime/cyborg_agent.py

@@ -0,0 +1,10 @@
+from CybORG.Agents import RandomAgent
+
+
+class MyAgent(RandomAgent):
+    """Baseline CybORG blue-team agent.
+
+    Improve this class to choose better defensive actions in the simulated DroneSwarm scenario.
+    """
+
+    pass

codeclash/arenas/cyborg/runtime/run_cyborg.py

@@ -0,0 +1,155 @@
+import argparse
+import importlib.util
+import json
+import random
+import re
+import traceback
+from pathlib import Path
+from statistics import mean
+
+import numpy as np
+from CybORG import CybORG
+from CybORG.Agents import BaseAgent
+from CybORG.Agents.Wrappers.PettingZooParallelWrapper import PettingZooParallelWrapper
+from CybORG.Simulator.Scenarios import DroneSwarmScenarioGenerator
+
+CRASH_SCORE = -1_000_000.0
+
+
+def safe_module_name(player_name: str) -> str:
+    safe = re.sub(r"\W+", "_", player_name)
+    if not safe or safe[0].isdigit():
+        safe = f"player_{safe}"
+    return f"codeclash_cyborg_{safe.lower()}"
+
+
+def load_agent_class(player_name: str, path: str):
+    spec = importlib.util.spec_from_file_location(safe_module_name(player_name), path)
+    if spec is None or spec.loader is None:
+        raise RuntimeError(f"Could not load module spec from {path}")
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    if not hasattr(module, "MyAgent"):
+        raise RuntimeError(f"{path} does not define MyAgent")
+    agent_class = module.MyAgent
+    if not issubclass(agent_class, BaseAgent):
+        raise RuntimeError(f"{path} MyAgent must inherit from CybORG BaseAgent")
+    return agent_class
+
+
+def make_agent(agent_class: type, agent_name: str):
+    try:
+        return agent_class(name=agent_name)
+    except TypeError:
+        try:
+            return agent_class(agent_name)
+        except TypeError:
+            return agent_class()
+
+
+def evaluate_player(
+    player_name: str,
+    agent_class: type,
+    *,
+    episode_idx: int,
+    steps: int,
+    drones: int,
+) -> dict:
+    seed = 4100 + episode_idx
+    random.seed(seed)
+    np.random.seed(seed)
+
+    try:
+        scenario = DroneSwarmScenarioGenerator(num_drones=drones)
+        env = PettingZooParallelWrapper(CybORG(scenario, "sim"))
+        observations = env.reset()
+        action_spaces = env.action_spaces
+        agents = {agent_name: make_agent(agent_class, agent_name) for agent_name in env.possible_agents}
+
+        for agent_name, agent in agents.items():
+            if hasattr(agent, "set_initial_values"):
+                agent.set_initial_values(action_spaces[agent_name], observations[agent_name])
+
+        step_rewards = []
+        for _ in range(steps):
+            actions = {
+                agent_name: agents[agent_name].get_action(observations[agent_name], action_spaces[agent_name])
+                for agent_name in env.agents
+            }
+            observations, rewards, done, _info = env.step(actions)
+            step_rewards.append(mean(rewards.values()))
+            if all(done.values()):
+                break
+
+        for agent in agents.values():
+            if hasattr(agent, "end_episode"):
+                agent.end_episode()
+
+        return {
+            "player": player_name,
+            "episode": episode_idx,
+            "score": float(sum(step_rewards)),
+            "steps_completed": len(step_rewards),
+            "status": "ok",
+        }
+    except Exception as exc:
+        return {
+            "player": player_name,
+            "episode": episode_idx,
+            "score": CRASH_SCORE,
+            "steps_completed": 0,
+            "status": "error",
+            "error": f"{type(exc).__name__}: {exc}",
+            "traceback": traceback.format_exc(limit=5),
+        }
+
+
+def parse_agent_arg(value: str) -> tuple[str, str]:
+    if "=" not in value:
+        raise argparse.ArgumentTypeError("--agent values must be NAME=/path/to/cyborg_agent.py")
+    name, path = value.split("=", 1)
+    if not name:
+        raise argparse.ArgumentTypeError("agent name cannot be empty")
+    if not Path(path).exists():
+        raise argparse.ArgumentTypeError(f"agent path does not exist: {path}")
+    return name, path
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--agent", action="append", type=parse_agent_arg, required=True)
+    parser.add_argument("--episodes", type=int, default=3)
+    parser.add_argument("--steps", type=int, default=30)
+    parser.add_argument("--drones", type=int, default=18)
+    parser.add_argument("--output", required=True)
+    args = parser.parse_args()
+
+    agent_classes = {name: load_agent_class(name, path) for name, path in args.agent}
+    totals = {name: 0.0 for name in agent_classes}
+    details = []
+
+    for episode_idx in range(args.episodes):
+        for player_name, agent_class in agent_classes.items():
+            result = evaluate_player(
+                player_name,
+                agent_class,
+                episode_idx=episode_idx,
+                steps=args.steps,
+                drones=args.drones,
+            )
+            totals[player_name] += result["score"]
+            details.append(result)
+
+    averages = {player: score / args.episodes for player, score in totals.items()}
+    output = {
+        "average_scores": averages,
+        "total_scores": totals,
+        "episodes": args.episodes,
+        "details": [json.dumps(item, sort_keys=True) for item in details],
+    }
+    Path(args.output).parent.mkdir(parents=True, exist_ok=True)
+    Path(args.output).write_text(json.dumps(output, indent=2, sort_keys=True))
+
+
+if __name__ == "__main__":
+    main()