Radar / AI security / AWS SafeBox
AWS SafeBox — deterministic local guardrails for AI agents running infrastructure commands
A policy proxy that sits between an AI coding agent and tools like Terraform, AWS CLI, and kubectl; TOML rules block destructive commands (terraform destroy, recursive S3 deletes, instance termination) before the real executable runs, failing closed on anything it cannot evaluate.
Why it matters
Deterministic allow/block by explicit policy instead of asking an LLM whether a command looks safe — a seatbelt layer for autonomous infra agents.
What you could build with it
A DevOps startup could wrap its AI coding agents in a deterministic command proxy so overnight autonomous runs can plan and apply infrastructure changes without a human babysitting every shell invocation, while commands like terraform destroy are always blocked.
Does it hold up?
Evidence thin but concrete — the README documents manual verification on macOS plus Linux-container testing and 59 passing tests; the authors are explicit about limitations (PATH-proxy bypassable by absolute paths, not a substitute for IAM/SCPs), and no independent adoption reports exist yet.
Built with AWS SafeBox
- aws-safebox/aws-safeboxgithub · 2 ★ · Community Edition 0.1.0: policy engine, executable proxy, custom TOML policies, CLI (check/run/setup).
Learn more
First spotted on github: source.
More AI security
iFixAi: independent auditing of AI agentsIndependent AI-agent auditor: 250+ inspections across 69 misalignment categories (red teaming + operational assurance)…security · JEV 0.76OpenAI, Anthropic and Google DeepMind jointly unveil cyber-focused safety models and safeguardsThe three rival labs disclosed weeks of behind-the-scenes coordination and jointly released a set of cyber-focused AI…security · JEV 0.73OpenAI textGrain watermarking for ChatGPT and Codex in the EUOpenAI will roll out its invisible textGrain watermarking system to ChatGPT/Codex users in the EU over the coming…security · JEV 0.65Cloudflare security-audit-skill: multi-phase security audits for coding agentsA coding-agent skill that runs multi-phase security audits with independently verified, machine-checkable findings.security · JEV 0.63LiveNerf: a pre-registered benchmark for post-release model driftOpen-source, pre-registered 30-day benchmark that detects whether Claude Opus 5.5 quietly gets worse after launch.security · JEV 0.61Sandlock 0.8.8: deferred commit for agent sandboxesThe process-based Linux AI sandbox (no container, no VM) ships deferred commit: every run returns a changeset of what…security · JEV 0.58
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.