An agent polling protocol is the request cycle a build agent uses to contact its server and ask for work, updates, or configuration. In TeamCity, this channel lets agents regularly check in so the server can assign jobs and coordinate execution. It is part of the control plane, not the build output itself.
This matters in cyber security because polling traffic often sits close to high-value orchestration logic, credentials, and build secrets. If an attacker can tamper with this path, they may be able to influence job assignment, extract sensitive data, or reach code execution in the server process. Defenders protect it by limiting network exposure, enforcing authentication and authorization, separating servers from agents where possible, and patching quickly when protocol handling flaws are found.