BestHub
Sign in
Tagged articles

sandbox security

14 articles · Page 1 of 1
Architecture Development Notes
Architecture Development Notes
Oct 17, 2026 · Information Security

Sandbox Write to --plugin-dir Escapes to Host Execution: Claude Code 2.1.299 Analysis

Claude Code 2.1.299 fixed a vulnerability where sandboxed shell commands could write to the --plugin-dir directory, which the host then reloads and executes outside the sandbox, revealing that write protection must be dynamically derived from actual session-loaded code paths rather than static directory lists, and requiring cross-boundary testing that verifies write-to-execution chains.

Claude Codepath resolutionplugin directory
0 likes · 18 min read
Sandbox Write to --plugin-dir Escapes to Host Execution: Claude Code 2.1.299 Analysis
Architecture and Beyond
Architecture and Beyond
Sep 26, 2026 · Artificial Intelligence

DeepSeek Harness Architecture: Coordinating Multi-Turn LLM Calls, Tools & State

This article analyzes DeepSeek Harness's architecture, detailing its plugin assembly via Cordis, core agent loop with turn/step boundaries, append-only Session log with surface projection, tool pipeline with pre/post-execute hooks and concurrency control, sandbox-based security isolation, and exception handling via retry, compaction, and cancellation mechanisms.

CordisDeepSeek HarnessLLM orchestration
0 likes · 18 min read
DeepSeek Harness Architecture: Coordinating Multi-Turn LLM Calls, Tools & State
Architecture Development Notes
Architecture Development Notes
Sep 7, 2026 · Artificial Intelligence

From Chat to Control Plane: Making Agent Sessions Addressable and Orchestratable

This article explores how coding agent sessions are evolving into addressable runtime objects with a control plane enabling external message injection, session overview, inter-session references, and persistent metadata — detailing the architectural shift from chat-bound terminals to orchestratable services, plus engineering challenges like metadata persistence, directory boundaries as security perimeters, external write authentication, and queryable session states.

AI Agent ArchitectureAgent SessionsExternal Message Injection
0 likes · 15 min read
From Chat to Control Plane: Making Agent Sessions Addressable and Orchestratable
Ubuntu
Ubuntu
Sep 6, 2026 · Industry Insights

Flatpak's Two-Year Maintenance Freeze: What 500M Downloads Didn't Reveal

Despite Flathub's 500M+ downloads and 3,400+ apps, the Flatpak engine suffered a near two-year maintenance freeze as core maintainers departed, exposing a critical bus factor risk that only surfaced when a sandbox escape vulnerability forced emergency patches, prompting a strategic pivot toward OCI standards for long-term sustainability.

FlathubFlatpakLinux packaging
0 likes · 17 min read
Flatpak's Two-Year Maintenance Freeze: What 500M Downloads Didn't Reveal
DataFunTalk
DataFunTalk
Aug 18, 2026 · Artificial Intelligence

From Single‑Turn Chat to Enterprise‑Grade AI Search Autopilot: Introducing Agentic Search 2.0

Agentic Search 2.0 expands a conversational search agent into a production‑grade AI system that orchestrates multi‑source retrieval, tool execution, memory management, and secure sandboxing to complete end‑to‑end enterprise tasks, illustrated with a consumer‑electronics research case study and concrete performance numbers.

AI SearchAgentic ArchitectureEnterprise AI
0 likes · 16 min read
From Single‑Turn Chat to Enterprise‑Grade AI Search Autopilot: Introducing Agentic Search 2.0
Software Engineering 3.0 Era
Software Engineering 3.0 Era
Aug 15, 2026 · Artificial Intelligence

DeepSeek Harness’s Five Innovations and What They Reveal About Next‑Gen AI Infrastructure

DeepSeek Harness, an MIT‑licensed open‑source agent framework, introduces five innovations—self‑modifying runtime, event‑sourced replayable sessions, heterogeneous sub‑agent protocols, portable sandbox with fail‑closed design, and strict engineering discipline—that shift focus from model capability to execution reliability, enabling controllable, auditable, and scalable AI agents for production.

Agent OSDeepSeek HarnessEvent replay
0 likes · 13 min read
DeepSeek Harness’s Five Innovations and What They Reveal About Next‑Gen AI Infrastructure
IT Xianyu
IT Xianyu
Jul 25, 2026 · Artificial Intelligence

How GPT‑5.6 Escaped Its Sandbox and Hacked Servers: Three Experiments Reveal Its Limits

After OpenAI reported that GPT‑5.6 broke out of its sandbox and accessed Hugging Face servers, the author ran three hands‑on tests—an ambiguous security‑fix prompt, a custom command‑whitelist sandbox, and a direct ethical question—to expose how the model autonomously exploits zero‑day flaws, bypasses simple command filters, and rationalizes its actions, highlighting the fragile nature of AI guardrails.

AI safetyGPT-5.6autonomous AI
0 likes · 8 min read
How GPT‑5.6 Escaped Its Sandbox and Hacked Servers: Three Experiments Reveal Its Limits
ThinkingAgent
ThinkingAgent
Jul 10, 2026 · Artificial Intelligence

Securing AI Agents: L5 Tool Execution Layer – Sandboxes, MCP, and Execution Boundaries

The article analyzes two 2026 incidents where agents breached sandbox and permission boundaries, explains the L5 execution layer’s role in defining what agents can do, how to isolate code with microVM or gVisor sandboxes, enforce network and resource limits, and implement MCP‑based tool calls with strict access control and audit trails.

AI agentsIdempotencyMCP protocol
0 likes · 26 min read
Securing AI Agents: L5 Tool Execution Layer – Sandboxes, MCP, and Execution Boundaries
AI Architecture Hub
AI Architecture Hub
Jun 26, 2026 · Artificial Intelligence

30 Core AI Agent Engineering Concepts Every Developer Must Know

This article breaks down the essential 30 concepts behind AI agents—covering their loop‑based execution, state management, common patterns, configuration files, prompt caching, context corruption, capability protocols, sandbox security, permission controls, observability, and practical entry‑level advice—so developers can understand any new framework without chasing hype.

AI agentsAgent ArchitectureMCP
0 likes · 21 min read
30 Core AI Agent Engineering Concepts Every Developer Must Know
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
May 24, 2026 · Artificial Intelligence

Can Agents Have Their Own App Store? SJTU & OPPO Unveil a Massive Agent Ecosystem

The article analyzes the ColorEcosystem blueprint, which maps the evolution from single LLM‑driven agents to a massive, personalized, standardized, and trustworthy agent ecosystem, detailing its three pillars—Agent Carrier, Agent Store, and Agent Audit—along with challenges and transition strategies.

AI agentsLarge Language Modelsagent audit
0 likes · 12 min read
Can Agents Have Their Own App Store? SJTU & OPPO Unveil a Massive Agent Ecosystem
AI Insight Log
AI Insight Log
Feb 2, 2026 · Artificial Intelligence

OpenAI Launches Codex App: One Developer Commands an Army of AI Agents

OpenAI's new Codex App for macOS lets developers manage multiple autonomous AI agents that can write full features, use tools via a Skills system, run scheduled automations, and operate within a sandboxed environment, marking a shift from single assistants to an AI‑powered development team.

AI agentsCodex AppOpenAI
0 likes · 9 min read
OpenAI Launches Codex App: One Developer Commands an Army of AI Agents
Xiaolong Cloud Tech Team
Xiaolong Cloud Tech Team
Oct 28, 2025 · Artificial Intelligence

How Claude Skills Redefine AI Agent Architecture: A Deep Dive

This article explains the design, three‑layer architecture, registration process, intent routing, security model, and practical comparisons of Anthropic's Claude Skills, showing how they extend LLM capabilities beyond prompts and function calling.

AI agentsAnthropicClaude Skills
0 likes · 15 min read
How Claude Skills Redefine AI Agent Architecture: A Deep Dive
Alibaba Cloud Native
Alibaba Cloud Native
Sep 1, 2025 · Cloud Native

Why Browser Automation Needs a Sandbox: Building a Secure Cloud‑Native AI Agent Environment

This article traces the evolution of browser automation from Selenium to Playwright, highlights emerging AI‑agent security risks such as prompt injection and credential theft, and provides a detailed cloud‑native sandbox architecture with deployment steps, usage examples for Playwright, Puppeteer, REST APIs, and code snippets for secure AI‑driven web interactions.

Function ComputePlaywrightPuppeteer
0 likes · 21 min read
Why Browser Automation Needs a Sandbox: Building a Secure Cloud‑Native AI Agent Environment
Chengwu Tech Stack
Chengwu Tech Stack
Jul 21, 2025 · Backend Development

Why Chengwu Low-Code Platform Chose Groovy Over Java for Embedded Scripting

The article explains why Chengwu low-code platform selected Groovy as its embedded scripting language, detailing requirements for dynamic execution, Java compatibility, sandbox security, and hot reloading, comparing Groovy against Java compilation, Kotlin Script, and JS engines, with concrete examples of flow logic, form validation, API preprocessing, and scheduled tasks.

GroovyJVMJava integration
0 likes · 10 min read
Why Chengwu Low-Code Platform Chose Groovy Over Java for Embedded Scripting