OPENMOAT BLOG
How it works, how to use it, and what we learn building it.
Guides and explainers on keeping AI coding agents away from your secrets, your files and your network. RSS
One set of rules for Claude Code, Codex and Cursor on your team
Your team uses three agents with three settings formats. How a committed .moat/policy.yaml gives everyone the same deny and ask rules, why its allow rules need trust, and how to verify the audit log.
What a hook can't see inside npm test
A hook sees "npm test" and allows it. The script inside can read your keys. How the agent's OS sandbox closes that gap, the three sandbox tiers, and the published test results.
Keep .env files and API tokens away from your AI agent
Which secrets the agent can never read, the ones it can, and how a brokered token lets it call an API with a placeholder instead of the real value.
Your agent's prompt shows the command, not what it does
Variables, base64, globs, quoting tricks and commands inside commands. Why reading the command text is not enough, and what OpenMoat checks instead.
What still stops a bad command when your agent runs without prompts
--dangerously-skip-permissions, Run Everything, CI and background agents: what still stops a bad command when nobody answers a prompt.
Introducing OpenMoat
What OpenMoat is, the problem it solves, how it decides, what it protects, where it stops, and how to try it in two commands.
No posts with this tag yet.