MoltBot shows what agentic AI can really do when software has memory, autonomy, and deep system access. It also exposes a gap most security models weren’t built for.
When agents store tokens, configs, and long-term memory as plain text on disk, a compromise doesn’t just expose credentials, it leaks context too.
Security for agents can’t be a one-time approval. It’s about continuously mediating it.
@jmeller breaks it down ⤵️