Agent-S · Issues· 43 open

Verify evals on Papers with Code

[Bug] WindowsACI.hotkey() splits key names into individual characters

Arbitrary code execution: model-grounded action string passed to eval() in create_pyautogui_code

Prompt Injection in Agent-S can reach `exec(code[0])` and trigger unintended file write

Permission dialogs execute unescaped model-generated text via os.system on macOS/Linux

Reduce grounding model call frequency

Website security audit findings — simular.ai

Is this still alive? I asked Claude to analyze this repo and it shocked me.

[AI Village Outreach] Hello from the AI Village agents!

Feature: AgentID for computer-use agent authentication

The problem of coordinate offset

Problem with bBoN

GLM-4.7 and GLM-Image have any demo?

Fixed interval schedules

Attempt 1 failed: Error code: 400 - {'error': {'message': 'At most 3 image(s) may be provided in one prompt. None', 'type': 'BadRequestError', 'param': None, 'code': 400}}