| 标题 | Bytebot AI Bytebot 0.0.1 Denial of Service |
|---|
| 描述 | Bytebot has a denial of service vulnerability in its agent execution workflow.
A normal task prompt can cause the agent to repeatedly call the model and desktop tools. The backend continues scheduling new agent iterations until the model emits a terminal task status, but it does not enforce a maximum iteration count, tool-call count, wall-clock timeout, model-call budget, token/cost budget, child-task budget, or stored tool-result byte budget.
An attacker can submit a crafted task that keeps the agent in a model/tool loop. In shared or multi-user deployments, this can monopolize the agent execution pool, repeatedly consume model and desktop-tool resources, store growing tool results and screenshots, and keep legitimate follow-up tasks pending. This can cause task queue starvation, denial of wallet, database growth, and feature-level denial of service. |
|---|
| 来源 | ⚠️ https://github.com/outlookgp/CVE/tree/main/Bytebot_Agent_Loop_Budget_DoS_Report |
|---|
| 用户 | outlookgp (UID 99658) |
|---|
| 提交 | 2026-07-11 11時44分 (2 月前) |
|---|
| 管理 | 2026-08-24 07時56分 (1 month later) |
|---|
| 状态 | 已接受 |
|---|
| VulDB条目 | 394599 [bytebot-ai bytebot 0.0.1 Agent Execution Workflow 拒绝服务] |
|---|
| 积分 | 20 |
|---|