From RMM Agent to Remote Shell: Tracing the N-central Zero-Day Chains
N-able took N-central offline on 14 November 2026 after researchers published details on 33 zero-days, and a day later two exploit chains were named: Agent2Shell and 4EverAble2Shell. This short walks through what is actually reported, what is only inference, and the architecture that makes the story serious. An RMM agent already runs scripts as SYSTEM, pushes binaries, restarts services, writes config and opens remote control sessions. That is close to an attacker's wishlist. The interesting question is the trust boundary: the agent dials out, authenticates, and then asks the server what to do, so outbound-only describes network direction and not authority. Where the enrollment credential lives, whether it is static per tenant, and whether it can be transplanted are the questions worth testing in a lab. The blast radius is what gives the chains their names. One console session can reach every client an MSP manages, and stored "run as" references mean an RMM breach can become a domain breach. What is missing matters just as much: no CVE identifiers, no affected version range, no statement about authentication requirements, and no confirmation of in-the-wild exploitation in the material summarised here. Agent2Shell moving agent to interactive shell, and 4EverAble2Shell restoring revoked access, are readings of the names, not evidence. Practical steps covered: mirror console audit events, agent inventory and credential references outside the platform, because an outage also freezes audit evidence; treat long-lived enrollment tokens as shared passwords; and baseline the agent's outbound connections on an authorised host. Do not run speculative exploit attempts against any N-central server or agent, yours or a client's, since without a vendor advisory you cannot know whether a probe is harmless. Full write-up: https://hackyjs.com/posts/from-rmm-agent-to-remote-shell-tracing-the-n-central-zero-day-chains