SShortSingh.
Back to feed

Google Launches Agent Sandbox to Safely Run AI-Generated Code on Kubernetes

0
·1 views

Google Cloud has released Agent Sandbox, a managed GKE feature that provides isolated, stateful Linux containers for executing untrusted, LLM-generated code. Each sandbox runs as a single-replica Kubernetes workload with a stable hostname, persistent storage, kernel-level isolation, and a default deny-all network policy. The system is built on Kubernetes custom resource definitions, including primitives for sandbox instances, reusable templates, and warm pools that allow pre-started sandboxes to be assigned in milliseconds. An open-source controller under the kubernetes-sigs/agent-sandbox project lets teams run the same setup on their own clusters without relying on Google's managed offering. The feature addresses a growing security concern for platform engineers as AI agents are increasingly tasked with running shell commands, installing packages, and interacting with live infrastructure.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Anthropic Releases Claude Code 2.1.224 With Self-Hosted Runner Support for Enterprises

Anthropic released Claude Code 2.1.224 on August 7, 2026, introducing self-hosted environment support for Team and Enterprise organizations. The update allows a new claude self-hosted-runner process to execute Claude Code cloud sessions on an organization's own machines or containers. However, the feature does not constitute fully on-premises inference — repository checkouts, build artifacts, and injected secrets remain on the organization's infrastructure, while session prompts, responses, transcripts, and model inference continue to flow through Anthropic's control plane at api.anthropic.com. The public beta currently excludes Zero Data Retention organizations, non-GitHub repositories, and inference routing through Bedrock, Google Cloud, or Microsoft Foundry. Platform teams are advised to carefully evaluate the trust boundary and complete a seven-step security checklist before connecting any production repository to a self-hosted runner.

0
ProgrammingDEV Community ·

Developer Builds Offline, Encrypted Calibration Tool for Small Factories Using .NET 8

A developer created CalibKeep, a fully offline desktop application designed to replace spreadsheet-based calibration tracking in small manufacturing facilities. Built with .NET 8, Avalonia UI, and an AES-256 encrypted local database via SQLCipher, the app manages calibration schedules, depreciation, and audit trails. A critical bug was discovered where opening a SQLCipher database with the wrong passphrase permanently corrupted the connection handle, which was resolved by verifying an HMAC marker before ever accessing the SQLite file. The app supports ISO 9001 compliance by writing audit log entries within the same transaction for every data change, and a rule engine flags high-impact changes when calibration-critical fields are modified. The tool was published to the Microsoft Store, where the developer encountered and resolved issues around app manifest formatting, certificate signing mismatches, and license validation logic.

0
ProgrammingDEV Community ·

Tech Industry's Title Inflation Is Misleading Hiring Managers and Hurting Candidates

A hiring manager's policy of auto-rejecting Staff or Principal engineers with only 3-4 years of experience ignited debate online, with critics split between calling it gatekeeping and calling it reasonable. Research shows the controversy stems from widespread title inflation, with studies indicating that a quarter of junior roles were rebranded with senior titles between 2019 and 2022, while Lead positions tripled among early-career workers by 2023. A Pearl Meyer survey of over 400 organizations found 54% were using inflated titles as a talent attraction tool, often as a substitute for competitive pay raises. Experts like author Will Larson argue that Staff-level roles should be defined by scope and influence rather than tenure, meaning some engineers at small startups holding the title may not meet that bar, while others with fewer years genuinely do. Ultimately, candidates are the primary victims of this inflation, penalized by recruiters for titles they were given rather than ones they chose.

Google Launches Agent Sandbox to Safely Run AI-Generated Code on Kubernetes · ShortSingh