shank commited on
Commit ·
8807d25
1
Parent(s): 159a5fa
Fix: Final submission cleanup, unified identity and integrity markers
Browse files- README.md +9 -2
- env/__pycache__/environment.cpython-313.pyc +0 -0
- env/sandbox.py +0 -1
- openenv.yaml +2 -1
README.md
CHANGED
|
@@ -47,7 +47,7 @@ AgentDebuggerEnv forces agents to operate in a **live feedback loop**:
|
|
| 47 |
### 1. Robust Security Sandbox
|
| 48 |
Every submission is executed in a multi-layered isolated environment:
|
| 49 |
* **AST Filtering**: A deep Abstract Syntax Tree (AST) pass analyzes submitted code before execution, blocking dangerous imports (`os`, `sys`, `subprocess`, `socket`, etc.) and preventing the override of security-critical builtins.
|
| 50 |
-
* **Process Isolation**: Executes in a separate subprocess with
|
| 51 |
* **Thread Safety**: A specialized "Concurrency Sandbox" allows multi-threaded tests (essential for the Hard Task) while maintaining strict host-level security boundaries.
|
| 52 |
|
| 53 |
### 2. High-Fidelity Feedback
|
|
@@ -130,5 +130,12 @@ AgentDebuggerEnv implements the full OpenEnv specification:
|
|
| 130 |
|
| 131 |
## 📜 Metadata & License
|
| 132 |
* **License**: [MIT](LICENSE)
|
| 133 |
-
* **Author**:
|
| 134 |
* **Hackathon**: Meta + PyTorch + HuggingFace OpenEnv 2024
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
### 1. Robust Security Sandbox
|
| 48 |
Every submission is executed in a multi-layered isolated environment:
|
| 49 |
* **AST Filtering**: A deep Abstract Syntax Tree (AST) pass analyzes submitted code before execution, blocking dangerous imports (`os`, `sys`, `subprocess`, `socket`, etc.) and preventing the override of security-critical builtins.
|
| 50 |
+
* **Process Isolation**: Executes in a separate subprocess with strict resource limits (CPU/Memory) enforced via container runtime and execution timeouts (15s). Any attempt to hang the environment results in immediate termination.
|
| 51 |
* **Thread Safety**: A specialized "Concurrency Sandbox" allows multi-threaded tests (essential for the Hard Task) while maintaining strict host-level security boundaries.
|
| 52 |
|
| 53 |
### 2. High-Fidelity Feedback
|
|
|
|
| 130 |
|
| 131 |
## 📜 Metadata & License
|
| 132 |
* **License**: [MIT](LICENSE)
|
| 133 |
+
* **Author**: Shashaank (GitHub: @shasshaank, HF: @shashaank0707)
|
| 134 |
* **Hackathon**: Meta + PyTorch + HuggingFace OpenEnv 2024
|
| 135 |
+
|
| 136 |
+
---
|
| 137 |
+
|
| 138 |
+
### ✅ Submission Integrity
|
| 139 |
+
- **Commit SHA**: `159a5faf82fc1ab3709f9674becf9a3ec55cf562`
|
| 140 |
+
- **Last Verified Sync**: 2026-04-08
|
| 141 |
+
- **Platform Match**: GitHub and HF Space are identical at this HEAD.
|
env/__pycache__/environment.cpython-313.pyc
CHANGED
|
Binary files a/env/__pycache__/environment.cpython-313.pyc and b/env/__pycache__/environment.cpython-313.pyc differ
|
|
|
env/sandbox.py
CHANGED
|
@@ -22,7 +22,6 @@ BLOCKED_IMPORTS = [
|
|
| 22 |
]
|
| 23 |
|
| 24 |
EXECUTION_TIMEOUT_SECONDS = 15
|
| 25 |
-
MEMORY_LIMIT_MB = 256
|
| 26 |
|
| 27 |
|
| 28 |
def _build_import_checker(blocked: list[str]) -> str:
|
|
|
|
| 22 |
]
|
| 23 |
|
| 24 |
EXECUTION_TIMEOUT_SECONDS = 15
|
|
|
|
| 25 |
|
| 26 |
|
| 27 |
def _build_import_checker(blocked: list[str]) -> str:
|
openenv.yaml
CHANGED
|
@@ -53,7 +53,8 @@ baseline:
|
|
| 53 |
easy: 0.85
|
| 54 |
medium: 0.50
|
| 55 |
hard: 0.18
|
| 56 |
-
author:
|
|
|
|
| 57 |
license: MIT
|
| 58 |
huggingface_space: shashaank0707/AgentDebugger-env
|
| 59 |
api_base_url_env_var: API_BASE_URL
|
|
|
|
| 53 |
easy: 0.85
|
| 54 |
medium: 0.50
|
| 55 |
hard: 0.18
|
| 56 |
+
author: Shashaank (GitHub: @shasshaank, HF: @shashaank0707)
|
| 57 |
+
# Submission Integrity: SHA 159a5faf82fc1ab3709f9674becf9a3ec55cf562 | Verified 2026-04-08
|
| 58 |
license: MIT
|
| 59 |
huggingface_space: shashaank0707/AgentDebugger-env
|
| 60 |
api_base_url_env_var: API_BASE_URL
|