Hey everyone, we just open-sourced a new experimental project called Agent Vault. Short version: If you're running AI agents (Claude Code, Cursor, Codex, OpenClaw, Hermes etc.) with API keys in the environment, these keys are exposed to prompt injection. Agent Vault is a local proxy that injects credentials at the network boundary so the agent never actually sees them. Single Go binary, self-hosted, encrypted at rest. The agent makes API calls normally and the proxy handles auth on its behalf. This is separate from Infisical proper. Still a preview. We'd love for people to try it and tell us what's broken. Repo: https://github.com/Infisical/agent-vault Docs: https://docs.agent-vault.dev Happy to answer any questions here. Thanks! 😃